Anthropic's Claude AI branding displayed on a smartphone screenTrump's Ad-Hoc AI Vetting Regime Blocks and Revives Frontier Models
Intra-Party Split Detected
Trump administration officials and industry are split between the White House's traditional deregulatory, laissez-faire stance on AI and a more interventionist faction pushing export controls, mandatory vetting, and government stakes in AI companies, creating tension with libertarian-leaning conservatives and industry allies who favor light-touch oversight to beat China.
Left says
- •The absence of durable legislation forced a reactive, personality-driven process where a single Amazon-flagged jailbreaking report triggered sweeping export controls and a three-week global blackout of frontier models, showing how fragile ad-hoc governance is.
- •Congress and the Biden administration's earlier disclosure requirements, which Trump scrapped, might have prevented this chaos entirely by giving regulators and companies a shared framework for judging severity before crises erupt.
- •The scramble reveals a government under-resourced to vet AI, with agencies like CAISI critically underfunded relative to what Trump's own AI action plan requires, raising doubts about whether officials have the technical capacity to make sound safety calls.
- •OpenAI's proposal to give the government an equity stake alongside the same administration deciding when its models can launch raises serious conflict-of-interest concerns, since regulators would have a financial incentive to wave models through.
Right says
- •The administration moved decisively to protect national security once credible jailbreaking vulnerabilities were identified, and Commerce Secretary Lutnick's swift action shows the government can act fast when stakes are high without waiting on Congress.
- •Trump's broader posture remains deregulatory, and officials describe the current cybersecurity executive order framework as voluntary, preserving flexibility and speed that companies like OpenAI say has been a genuinely productive collaboration.
- •Keeping frontier AI capabilities close to the vest, including limiting allied and adversarial access, is a reasonable extension of an America-first strategy to maintain technological and military dominance over China.
- •A potential government equity stake in OpenAI could give ordinary Americans direct financial exposure to AI's upside, mirroring the profitable Intel stake precedent, rather than letting all the gains accrue only to insiders and investors.
Common Take
High Consensus- Both sides agree the current model-vetting process is ad hoc and lacks a clear, standardized framework, even as Anthropic's Fable and Mythos models were restored and OpenAI's GPT-5.6/Sol went through government consultation before release.
- Both sides acknowledge that Amazon's jailbreaking report on Anthropic's models triggered real export controls and a weeks-long shutdown that industry, government, and outside cybersecurity experts all had to scramble to resolve.
- Both sides recognize that frontier AI capabilities have advanced dramatically in a short period, raising legitimate new questions about safety, national security, and how to compete with China's rapidly improving open-source models.
- Both sides want more clarity and a repeatable process going forward, as reflected in OpenAI's Sam Altman calling the current approach imperfect and lawmakers like Rep. Josh Gottheimer criticizing the confusion in the White House's vetting process.
The Arguments
Left argues
The absence of durable legislation meant a single disputed Amazon report could trigger sweeping export controls and a three-week global blackout of frontier models, demonstrating how fragile and unpredictable ad-hoc governance is when there's no agreed threshold for what counts as a dangerous vulnerability.
Right counters
The speed of the response is precisely the point: when Amazon flagged a credible jailbreaking risk, Lutnick and the administration moved within hours rather than waiting months for rulemaking, proving the government can act decisively on genuine national security threats without pre-existing legislation.
Right argues
Trump's posture remains fundamentally deregulatory and voluntary — officials describe the cybersecurity executive order as non-mandatory, and Sam Altman himself called the process 'productive,' suggesting industry and government found a workable, flexible arrangement rather than a heavy-handed regime.
Left counters
Calling it 'voluntary' is misleading when the alternative to compliance was a three-week global shutdown of a company's flagship product; that kind of coercive leverage without codified standards or appeal process is regulation by another name, just without accountability or predictability.
Left argues
CAISI's $15 million budget against an estimated $84 million requirement, combined with less than 1% of AI PhDs entering government, raises real doubts about whether the officials making launch-or-block decisions on frontier models have the technical depth to get those calls right.
Right counters
Even with resource constraints, the government successfully looped in NSA, CAISI, and multiple agencies to scrutinize Anthropic's models line-by-line before clearing them, showing that existing expertise — however thin — was sufficient to conduct a real technical review rather than a rubber-stamp.
Right argues
Restricting frontier AI access for adversaries and even scrutinizing allied access is a rational extension of America-first strategy, since AI is now integrated into military and intelligence capability and maintaining a technological edge over China is a legitimate national security priority.
Left counters
Alienating allies like the EU by blocking their access to top models while simultaneously asking them to join supply-chain pacts like Pax Silica risks fracturing the very coalition needed to out-compete China, since Europe may simply build parallel capabilities rather than remain a junior partner.
Right argues
A government equity stake in OpenAI, modeled on the profitable Intel precedent, could give ordinary Americans direct financial upside in the AI boom rather than letting all the gains flow to insiders, venture investors, and executives.
Left counters
Handing the government a financial stake in the very companies it's supposed to be regulating for safety creates an unavoidable conflict of interest — regulators who profit when a model launches have every incentive to wave through releases they might otherwise scrutinize, undermining the entire vetting process's credibility.
Challenge Questions
These questions target genuine internal contradictions — meant to provoke honest reflection.
Right asks Left
“If Biden's disclosure framework was scrapped and Congress still hasn't passed legislation, isn't the left's preferred solution just as speculative and untested as the ad-hoc regime it criticizes — how confident can anyone be that a 'durable framework' would have handled a novel jailbreaking report any better or faster?”
Left asks Right
“If the administration insists its cybersecurity framework is merely 'voluntary,' how does it square that claim with Anthropic having no real choice but to comply once Lutnick's letter effectively forced a global shutdown of its product — and doesn't that same coercive leverage undercut the deregulatory, hands-off image the administration wants to project to allies and industry?”
Outlier Report
Left Fringe
AI-safety accelerationist critics like some effective-altruism-adjacent commentators and figures such as Gary Marcus who want much stricter, codified regulation beyond what's described represent maybe 15-20% of the left, pushing for far more aggressive government control than mainstream Democrats propose.
Right Fringe
Tech-libertarian figures like Marc Andreessen and some America First accelerationists who oppose any export controls or government vetting at all, viewing even Trump's ad-hoc restrictions as excessive interference, represent roughly 15-20% of the right.
Noise Assessment
High noise ratio: most public discourse is driven by AI insiders, tech journalists, and policy wonks on X/Twitter (e.g., accounts like @sama, @petergostev) whose engagement dwarfs actual average-American attention to this niche regulatory story, which most ordinary Americans have not closely followed.
Sources (9)
<p>After years of touting ambitious climate goals, big tech companies are now facing new scrutiny over what they disclose about AI's environmental footprint.</p><p><strong>Why it matters:</strong> The AI boom is turning a handful of tech companies into defining players in the debate over electricity and water use. Their willingness to disclose these impacts is becoming almost as important as the impacts themselves.</p><hr /><p><strong>Driving the news</strong>: New environmental reports from <a href="https://www.axios.com/2026/06/30/google-ai-boom-emissions-power-use" target="_blank">Google</a>, <a href="https://www.axios.com/2026/07/01/amazon-ai-emissions-data-centers" target="_blank">Amazon</a> and <a href="https://www.axios.com/2026/07/09/microsoft-ai-data-centers-climate-goals" target="_blank">Microsoft</a> show emissions and water use continuing to rise as AI infrastructure expands — while revealing differences in what the companies disclose.</p><ul><li>"There is a bit of a reluctance to share a lot of things in this competitive dynamic and also with these being public companies," said Boris Gamazaychikov, who co-founded<strong> </strong>Sustainable AI Group, a new research and advisory firm that helps companies address the environmental impacts of AI.</li></ul><p><strong>The big picture</strong>: Transparency is emerging as a key response to <a href="https://www.axios.com/2026/06/22/ai-data-center-backlash-poll" target="_blank">growing opposition</a> to AI focused on data centers' energy and water toll.</p><ul><li>Amazon, Google, Microsoft and Meta account for <a href="https://www.axios.com/2026/06/26/amazon-google-microsoft-dominate-data-centers" target="_blank">roughly two thirds</a> of the data-center power capacity in a top 15 ranking by financial firm Jefferies, so what they do influences the entire market.</li></ul><p><strong>Friction point:</strong> United Nations Secretary-General António Guterres this week called on tech companies to publicly disclose the "full footprint" of their data centers, including carbon, water and land use.</p><ul><li>"AI may feel intangible — but its footprint is not," <a href="https://news.un.org/en/story/2026/07/1167873" target="_blank">Guterres said</a>, reiterating the U.N.'s AI Environmental Transparency Initiative <a href="https://apnews.com/article/climate-united-nations-london-guterres-cdba9edbe2081aec15b115d043b5f75c" target="_blank">unveiled</a> last month.</li></ul><p><strong>State of play: </strong>Top executives at these tech companies generally say they support transparency, but their reporting varies widely once you dig into the disclosures. </p><ul><li>No law requires companies to detail many of these AI-related environmental metrics, and no common reporting standard exists for those that are disclosed.</li></ul><p><strong>What they're saying: </strong>Kara Hurst, the chief sustainability officer at Amazon, said in an interview before the UN announcement that all companies building data centers should disclose their environmental footprints.</p><ul><li>"We'd love to see one holistic interoperable standard," Hurst said.</li></ul><p><strong>Catch up fast:</strong> Comparing the companies' disclosures is a complicated apples-to-oranges exercise.</p><ul><li>Alex de Vries-Gao, a researcher at VU Amsterdam university and founder of online platform Digiconomist, reviewed the latest disclosures and ranks them, on transparency alone, as Meta first, followed by Google and Microsoft in a "toss-up" for second, and Amazon last.</li><li>Meta's 2026 report is expected later this year, likely in the<strong> </strong>late summer or early fall.</li></ul><p><strong>Zoom in</strong>: Transparency doesn't necessarily equal top performance, and no company leads across every measure, according to de Vries-Gao's <a href="https://pdf.sciencedirectassets.com/776857/1-s2.0-S2666389925X00027/1-s2.0-S2666389925002788/main.pdf?X-Amz-Security-Token=IQoJb3JpZ2luX2VjEMH%2F%2F%2F%2F%2F%2F%2F%2F%2F%2FwEaCXVzLWVhc3QtMSJHMEUCIQD11RGLEqzCnW9zT8glhIzNdu42NjI2wQFZFdRPewsMHwIgHLNanCIpFyObrPlpb6VuwRdiiW3%2B90h8vNrYvmJy73gquwUIif%2F%2F%2F%2F%2F%2F%2F%2F%2F%2FARAFGgwwNTkwMDM1NDY4NjUiDDKZS2P7ayCYr2lCJSqPBYZb24NtQzRT3xBR6oVA6c1tmqT2xxGnKSJ5cYe6%2BLF4oHihNazgnsuZWO8y24iv2k0gI6zyIofqzbw6p%2FRt1voLZhzB2XpoT9I6F%2BdfhDYTycS2mpr9087Rh1%2BE9KXmwNXjCcLrEFLsON7bVDPGvGfUnGcEonLhGTLqrwqSw1iysk0ftdYDbh90rRJYHCVC%2BZ%2F1yTp5CCn6ibf2EnnJR23T2GOzhDU2ri8enaOYwcmbeOUinD4tzxd3BEfqwPOk9b8d4JGRcHmkNd%2B5RoY2tZG0YLXhcuhKRbjRtghtVVmG4aNBg0n2ofAim1xY3AUxu1xgVwQ8BiWuFHFJy4hbGrg89kMwlAvpDFgV1FGHkdCIC7LqbwTgzvazRpBJ5stv17moGPIHF5lYj9KGQYaddUxpP5efNE2HmtksIbf4zqiKB1TszhIG%2FG8sRT43pwW19RnzSqoU3nni%2BjHYqibNpV5cgAFxbDj10PG6Qn9EWuWELp3Qllgx62OGg46oc38iMMNWY90aXvIRlyvuZJwj8hO11pNKUTHbw3lB6elo8pJE0aybWc7txw4SdN5PjI6dO%2BDvWF9AoIe52lBGiSxCHYoMShle0%2BQMpsSeUL3ZpTf8oGRRuJGtp2h1KSzNNb8XH9VrGs%2FyxI9YxyUzpitzn2RI2uXPqibTTPkbB%2FSfEcWFADPe0h8LvKrcw%2BGtQTN982%2FVrEKGswNIE8ZHmrYMO8LhoKk%2FlXGW1HpReIwRrijI2gsAY2mMxoOmkczd%2Fr8haeHFh4Tgp7cNjHDsf0oxVnPW0JlaUuy%2FoVdSnu49WdjLMnOzj527ijG2apBkL2IF7061hNoREsLq%2B8UsJ3q9WZeRzqhL8%2F2IcCiyQ2KD2Z4w8vC50gY6sQFkybr50E8AZ5pWJuOr3FSB%2BBh7f2%2BQ2TtYnR%2FXMwdcCKPxuhRSDRBidlsxOQQo%2BRX81N0VnThoWcK2shTiAvuCoqt4W29BB7UeOLvQtUI2fQwecWDsxn1p7fNFn1HECMZNNqH1zhjU3MqnW%2BH0kG%2Bgddw7Z9PN5Sz%2FO1ZoRba9cPwG655zpnfR3n8yovvnhuNQ%2BUKdwRtMIq4eGklzbleJF3dRmy%2BGmm%2BJZm8C8a06FiE%3D&X-Amz-Algorithm=AWS4-HMAC-SHA256&X-Amz-Date=20260708T172035Z&X-Amz-SignedHeaders=host&X-Amz-Expires=300&X-Amz-Credential=ASIAQ3PHCVTYWAUPAYB6%2F20260708%2Fus-east-1%2Fs3%2Faws4_request&X-Amz-Signature=dc1898f74ba2dcb6d3e8634f8ea10240907743cffaaa21abfe09dc7f0221a9e6&hash=0e0e60e1396e7ef7b2ae3a6329feba14f000873ac1ffe4d7157ef42d72e0cb97&host=68042c943591013ac2b2430a89b270f6af2c76d8dfd086a07176afe7c76c2c61&pii=S2666389925002788&tid=spdf-45299b11-4958-4563-a37a-28130eabb28c&sid=f13589e32f2dd246480a1b5367f5073fce3cgxrqa&type=client&tsoh=d3d3LnNjaWVuY2VkaXJlY3QuY29t&rh=d3d3LnNjaWVuY2VkaXJlY3QuY29t&ua=0f15045159520c50050559&rr=a180c2ae98bda385&cc=us&kca=eyJrZXkiOiJXUlVUdHVqdFVkK01DVjRRRm9xV0tZbGQ5MmdaZUl1RHNOcUM2UDAzcmRESFBzVEdjdWVEbHh3aFFZUkF5OFQ3VWFiQ2w1OWdUcUJJUy9DZnRzMy9kclpwVXE2Q0xXdWZqZ2pYOCtaUVlFcitGMWdKN3MzUUtDU3l2ZktzM3FFNmtTMVZhMEh3Ry9lMjkvYnNJNURDZ0NVRGNPTHkvZ1RicXJudnhNenRiZ1pvUFA4VCIsIml2IjoiMjNhMDNiMmQ3MDBlZTRlODlmY2JmMGI4MDZhMTM5MzgifQ%3D%3D_1783531448842" target="_blank">analysis</a> published earlier this year and these latest reports. </p><ul><li>Google and Meta disclose the strongest energy-efficiency metrics, according to the analysis.</li><li>Microsoft newly disclosed specific water and power metrics at individual data center locations in its latest report, which de Vries-Gao said represents "a significant improvement in transparency."</li><li>Amazon reports the strongest water-efficiency figure, though it ranks last overall because it provides the fewest metrics, including how much electricity it consumes, de Vries-Gao said.</li><li>Google doesn't publish a companywide water-efficiency metric, arguing that a global average masks its local risk-based approach. Using Google's own data, de Vries-Gao estimates it would rank last.</li><li>Only Meta discloses the water associated with electricity generation — a potentially significant omission by the others<strong> </strong>because that use typically far exceeds water consumed at the data center itself.</li></ul><p><strong>Reality check:</strong> Disclosure is only one measure. Companies are also investing billions in clean-energy projects whose climate benefits could<strong> </strong>take years to materialize.</p><ul><li>On that front, Google is largely considered the leader, alongside Microsoft.</li></ul><p><strong>How it works:</strong> AI infrastructure needs both electricity and cooling. As data centers expand, companies are increasingly trading water-intensive cooling for more energy-intensive alternatives.</p><p><strong>Between the lines: </strong>Generating electricity from fossil fuels and nuclear power requires large amounts of water, and some experts say tech companies should account for that.</p><ul><li>Judging by Meta's 2025 disclosures, this indirect water use was roughly 24 times larger than the amount used at its data centers, de Vries-Gao <a href="https://pdf.sciencedirectassets.com/776857/1-s2.0-S2666389925X00027/1-s2.0-S2666389925002788/main.pdf?X-Amz-Security-Token=IQoJb3JpZ2luX2VjEMH%2F%2F%2F%2F%2F%2F%2F%2F%2F%2FwEaCXVzLWVhc3QtMSJHMEUCIQD11RGLEqzCnW9zT8glhIzNdu42NjI2wQFZFdRPewsMHwIgHLNanCIpFyObrPlpb6VuwRdiiW3%2B90h8vNrYvmJy73gquwUIif%2F%2F%2F%2F%2F%2F%2F%2F%2F%2FARAFGgwwNTkwMDM1NDY4NjUiDDKZS2P7ayCYr2lCJSqPBYZb24NtQzRT3xBR6oVA6c1tmqT2xxGnKSJ5cYe6%2BLF4oHihNazgnsuZWO8y24iv2k0gI6zyIofqzbw6p%2FRt1voLZhzB2XpoT9I6F%2BdfhDYTycS2mpr9087Rh1%2BE9KXmwNXjCcLrEFLsON7bVDPGvGfUnGcEonLhGTLqrwqSw1iysk0ftdYDbh90rRJYHCVC%2BZ%2F1yTp5CCn6ibf2EnnJR23T2GOzhDU2ri8enaOYwcmbeOUinD4tzxd3BEfqwPOk9b8d4JGRcHmkNd%2B5RoY2tZG0YLXhcuhKRbjRtghtVVmG4aNBg0n2ofAim1xY3AUxu1xgVwQ8BiWuFHFJy4hbGrg89kMwlAvpDFgV1FGHkdCIC7LqbwTgzvazRpBJ5stv17moGPIHF5lYj9KGQYaddUxpP5efNE2HmtksIbf4zqiKB1TszhIG%2FG8sRT43pwW19RnzSqoU3nni%2BjHYqibNpV5cgAFxbDj10PG6Qn9EWuWELp3Qllgx62OGg46oc38iMMNWY90aXvIRlyvuZJwj8hO11pNKUTHbw3lB6elo8pJE0aybWc7txw4SdN5PjI6dO%2BDvWF9AoIe52lBGiSxCHYoMShle0%2BQMpsSeUL3ZpTf8oGRRuJGtp2h1KSzNNb8XH9VrGs%2FyxI9YxyUzpitzn2RI2uXPqibTTPkbB%2FSfEcWFADPe0h8LvKrcw%2BGtQTN982%2FVrEKGswNIE8ZHmrYMO8LhoKk%2FlXGW1HpReIwRrijI2gsAY2mMxoOmkczd%2Fr8haeHFh4Tgp7cNjHDsf0oxVnPW0JlaUuy%2FoVdSnu49WdjLMnOzj527ijG2apBkL2IF7061hNoREsLq%2B8UsJ3q9WZeRzqhL8%2F2IcCiyQ2KD2Z4w8vC50gY6sQFkybr50E8AZ5pWJuOr3FSB%2BBh7f2%2BQ2TtYnR%2FXMwdcCKPxuhRSDRBidlsxOQQo%2BRX81N0VnThoWcK2shTiAvuCoqt4W29BB7UeOLvQtUI2fQwecWDsxn1p7fNFn1HECMZNNqH1zhjU3MqnW%2BH0kG%2Bgddw7Z9PN5Sz%2FO1ZoRba9cPwG655zpnfR3n8yovvnhuNQ%2BUKdwRtMIq4eGklzbleJF3dRmy%2BGmm%2BJZm8C8a06FiE%3D&X-Amz-Algorithm=AWS4-HMAC-SHA256&X-Amz-Date=20260708T172035Z&X-Amz-SignedHeaders=host&X-Amz-Expires=300&X-Amz-Credential=ASIAQ3PHCVTYWAUPAYB6%2F20260708%2Fus-east-1%2Fs3%2Faws4_request&X-Amz-Signature=dc1898f74ba2dcb6d3e8634f8ea10240907743cffaaa21abfe09dc7f0221a9e6&hash=0e0e60e1396e7ef7b2ae3a6329feba14f000873ac1ffe4d7157ef42d72e0cb97&host=68042c943591013ac2b2430a89b270f6af2c76d8dfd086a07176afe7c76c2c61&pii=S2666389925002788&tid=spdf-45299b11-4958-4563-a37a-28130eabb28c&sid=f13589e32f2dd246480a1b5367f5073fce3cgxrqa&type=client&tsoh=d3d3LnNjaWVuY2VkaXJlY3QuY29t&rh=d3d3LnNjaWVuY2VkaXJlY3QuY29t&ua=0f15045159520c50050559&rr=a180c2ae98bda385&cc=us&kca=eyJrZXkiOiJXUlVUdHVqdFVkK01DVjRRRm9xV0tZbGQ5MmdaZUl1RHNOcUM2UDAzcmRESFBzVEdjdWVEbHh3aFFZUkF5OFQ3VWFiQ2w1OWdUcUJJUy9DZnRzMy9kclpwVXE2Q0xXdWZqZ2pYOCtaUVlFcitGMWdKN3MzUUtDU3l2ZktzM3FFNmtTMVZhMEh3Ry9lMjkvYnNJNURDZ0NVRGNPTHkvZ1RicXJudnhNenRiZ1pvUFA4VCIsIml2IjoiMjNhMDNiMmQ3MDBlZTRlODlmY2JmMGI4MDZhMTM5MzgifQ%3D%3D_1783531448842" target="_blank">calculated</a>.</li><li>"We don't know whether that ratio is representative of the rest of the industry because the other companies don't disclose comparable figures," he said.</li></ul><p><strong>Not all environmental advocates </strong>agree this level of disclosure is warranted<strong>. </strong></p><ul><li>"It's probably not totally fair" to hold them accountable for water from associated electricity, said Peter Gleick, co-founder of the Pacific Institute, a California-based water research nonprofit. "But it's a legitimate question and would accelerate a push to switch to low-water using renewable energy sources."</li></ul><p><strong>What we're watching</strong>: Gamazaychikov, of Sustainable AI Group, says pressure from these tech companies' customers, like consumer brands, will likely prompt more transparency and accountability in the absence of government action.</p><p><strong>The bottom line: </strong>"I don't think anything<strong> </strong>regulatory is going to happen soon," he said. "It's going to take business to make this happen."</p>
<p><a href="https://www.axios.com/technology/automation-and-ai" target="_blank">AI</a> regulation in the U.S. over the last few months has been a frenzy. Some experts say it didn't have to be this way.</p><p><strong>Why it matters:</strong> AI companies and the government are praising each other for successful collaboration on rules for cutting-edge AI, but that view hides a scramble behind the scenes that could have been avoided.</p><hr /><p><strong>Driving the news</strong>: <a href="https://www.axios.com/2026/07/08/openai-gpt-trump-ban-lifted" target="_blank">OpenAI</a> and <a href="https://www.axios.com/2026/06/30/trump-anthropic-ai-model-fable-restrictions" target="_blank">Anthropic's</a> latest, most powerful models both ended up getting nods from the government before wide release. That would have been unthinkable just months ago.</p><p><strong>The big picture:</strong> The two companies now know first-hand what it's like to release powerful models under the Trump administration's approach to regulation.</p><ul><li>That's featured export control threats, licensing requirements and negotiations with a host of government agencies that are sometimes at odds. </li><li>Other AI labs are poised to face the same process, as a cybersecurity executive order detailing standards and procedures gets implemented. </li></ul><p><strong>Flashback: </strong>A <a href="https://www.axios.com/2023/10/30/ai-executive-order-biden-transparency-safety" target="_blank">Biden-era AI executive order</a> required companies to share safety testing results with the government, including whether their models could be tricked into bypassing built-in safety guardrails.</p><ul><li>The "jailbreaking" issue was the type of vulnerability Amazon flagged last month that eventually led to export controls on Anthropic.</li><li>President Trump, vowing to pursue a deregulatory agenda on AI, scrapped that order's reporting requirements.</li><li>When the Trump administration's safety concerns with Anthropic came to a head, there was no alignment with industry and government on how severe jailbreaks need to be to raise a red flag.</li><li>Had there been a framework to assess and standardize the severity of jailbreaking or safety bypassing, export controls may have been avoided, one source familiar with the situation told Axios.</li></ul><p><strong>What they're saying:</strong> The rest of the world has implemented tech privacy regulations, updated antitrust laws and passed transparency and research access measures, former Biden tech official Asad Ramzanali said.</p><ul><li>"We didn't do any of it," he said of the U.S., depriving the country of a strong foundation to create rules around AI today. </li><li>"Given where things are, this is the right thing for the companies and for the government," he added, referring to the Trump administration's efforts to set rules around powerful AI models. </li><li>"But we should never have been here." </li></ul><p><strong>The government failed to recruit and retain</strong> technical expertise from the outset, according to the Cato Institute's Kevin Frazier, noting <a href="https://hai.stanford.edu/assets/files/hai_ai-index-report-2024_chapter6.pdf" target="_blank">less than 1 percent</a> of AI Ph.D.s go into government.</p><ul><li>Agencies like the Cybersecurity and Infrastructure Security Agency and the Center for AI Standards and Innovation have also been "sidelined" and underfunded, said Frazier, who is also the director of the University of Texas' AI and law program.</li><li>CAISI's operational budget is $15 million but it needs $84 million annually to fulfill Trump's AI action plan, according to the <a href="https://ifp.org/funding-for-caisi/" target="_blank">Institute for Progress</a>.</li></ul><p><strong>"The truth is that there's never</strong> been a 'right' answer for how to govern AI, but there are plenty of wrong ones," including deficient government expertise and failing to build trust among the public, Frazier said.</p><ul><li>He argued that the government could have learned from state initiatives, pointing to examples like Oklahoma's free AI literacy training, Utah's regulatory sandboxes for testing AI under heightened oversight, and Massachusetts' data practices.</li></ul><p><strong>Meanwhile,</strong> <strong>Congress has failed</strong> to pass any comprehensive AI safety legislation that could help avoid these one-off regulatory fixes, despite years of bipartisan efforts and growing popular interest in reeling in AI.</p><ul><li>"Right now, there is far too much confusion with the White House's AI vetting process — both for the country and for our leading AI developers," Rep. Josh Gottheimer (D-N.J.) told Axios. </li></ul><p><strong>The other side: </strong>"I thought it was a very productive process," OpenAI CEO Sam Altman said on CNBC the day after the company announced it would be doing a wide release of GPT 5.6 following government negotiations.</p><ul><li>"This was our first time through it, so there are things we'll learn about how to make it better next time, which we'll get going on soon."</li></ul><p><strong>Between the lines:</strong> Former President Biden's use of the Defense Production Act to impose mandatory disclosure requirements on AI companies was <a href="https://www.axios.com/pro/tech-policy/2023/12/13/gop-digs-in-on-bidens-use-of-defense-powers-for-ai" target="_blank">viewed at the time</a> by Republicans and industry as an overreach but, under Trump, industry still faces significant pressure to collaborate.</p><ul><li>Trump has not invoked the DPA, and the administration says the provisions of his cyber executive order are voluntary.</li></ul><p><strong>But it's clear</strong> that in today's regulatory environment, with what happened to Anthropic's Fable fresh in everyone's minds, AI companies need to keep the government happy.</p><ul><li>"You really want to be confident in your safety claims because otherwise the world is going to get uncomfortable very fast," Altman said on CNBC.</li><li>The White House said Tuesday that they did not approve or disapprove OpenAI's decision to release a model.</li><li>"The top technology leaders in the country are fully lining up behind the President's commonsense approach," White House spokesperson Liz Huston said.</li></ul><p><strong>What we're watching:</strong> Industry and the administration are continuing to work on the voluntary framework required by the June AI executive <a href="https://www.axios.com/2026/06/02/trump-signs-new-ai-executive-order" target="_blank">order</a>. </p><ul><li>Per the order, it's due Aug. 1.</li></ul><p><em>Editor's note: This story has been updated to include a comment from the White House.</em></p>
<p>A staggering class divide now separates how Americans experience <a href="https://www.axios.com/technology/automation-and-ai" target="_blank">artificial intelligence</a>:</p><ul><li>For frontier power users, AI feels like a r<em>evolution</em>: a force capable of conjuring companies, building software and solving complex problems at warp speed.</li><li>For the average person, it feels more like an <em>evolution</em>: a smarter search bar, a faster inbox, an ambient tech layer that saves time — but not much else.</li></ul><p><strong>Why it matters: </strong>Trillions of dollars in economic value — and the livelihoods of <a href="https://www.axios.com/2026/05/27/ai-hype-doom-openai-anthropic" target="_blank">millions of workers</a> — are being staked on a technology that most Americans neither trust nor fully understand.</p><hr /><ul><li>It's a new chapter in America's <a href="https://www.axios.com/pro/tech-policy/2024/04/11/exclusive-poll-most-americans-see-the-digital-divide-as-an-election-issue" target="_blank">digital divide</a> — the AI "haves," "have-nots" and "know-nots" — with profound implications for the future of wealth, work and power.</li></ul><p><strong>Zoom in:</strong> The newest <a href="https://www.axios.com/2026/07/07/ai-hacking-benchmarking-tests" target="_blank">frontier models</a> are designed for an agentic world of coding, research and cybersecurity that most Americans will never see, let alone operate.</p><ul><li><a href="https://www.axios.com/2026/07/08/gpt-sol-ultra-openai-anthropic-grok" target="_blank">OpenAI's Sol</a> and <a href="https://www.axios.com/2026/06/27/anthropic-fable-5-return-soon" target="_blank">Anthropic's Fable</a> now sit atop the pyramid of elite AI obsession, prized for running long coding and research loops with minimal human intervention.</li><li>Prominent developers have spent the week personifying the two models — debating their temperaments, work ethics, even their personalities, the way sports fans argue over rival athletes.</li><li>"My overall feel is that Fable is a 'wise owl' who is very thoughtful and very well spoken," <a href="https://x.com/petergostev/status/2074918176354115886?s=20" target="_blank">tweeted</a> AI researcher Peter Gostev. "GPT-5.6-Sol is like a rottweiler who will grab the problem by the throat and not let go until it is done."</li></ul><p><strong>Reality check: </strong>The people fluent enough to judge Sol against Fable on a coding benchmark are a tiny slice of the country. For most Americans, those names and metrics mean nothing.</p><ul><li>Millions of people encounter AI passively or unknowingly — through search summaries, AI-generated content, customer-service bots and invisible features inside apps.</li><li><a href="https://www.pewresearch.org/internet/2026/06/17/americans-and-ai-2026-chatbots-smart-devices-and-views-on-impact/" target="_blank">Nearly half of U.S. adults</a> now use AI chatbots, but the most common use is basic information search — the same job Google has done for two decades, a world away from autonomous coding agents.</li><li>OpenAI counts more than 50 million paying subscribers in its weekly ChatGPT user base of more than 900 million. The population running agentic coding tools is a fraction of that fraction.</li></ul><p><strong>Between the lines: </strong>Even among the elites living the frontier AI revolution, there's a pecking order.</p><ul><li>Sol began as a <a href="https://www.axios.com/2026/06/25/trump-administration-openai-gpt-model-release" target="_blank">restricted preview</a> for OpenAI's trusted partners and select organizations before broader rollout, making early access itself a <a href="https://x.com/danshipper/status/2074859437374820691?s=20" target="_blank">status marker</a> inside AI circles.</li><li>Fable was <a href="https://www.axios.com/2026/06/13/anthropic-amazon-white-house" target="_blank">pulled offline globally</a> for nearly three weeks in June under U.S. export controls, while its more powerful sibling, <a href="https://www.axios.com/2026/04/08/mythos-system-card" target="_blank">Mythos</a>, remains restricted to a small number of trusted organizations.</li><li>The result is a hierarchy inside the hierarchy: free users, paid users, power users, preview users and an insider class testing capabilities the rest of the world can only read about.</li></ul><p><strong>The big picture: </strong>The AI industry ultimately needs broad social permission for the transformation it's selling: more data centers, deeper workplace automation, and AI embedded in schools, government and daily life.</p><ul><li>Yet as AI adoption has climbed, trust has fallen: 63% of Americans say AI is advancing too quickly, and just 16% expect it to benefit society over the next 20 years, according to <a href="https://www.pewresearch.org/internet/2026/06/17/americans-and-ai-2026-chatbots-smart-devices-and-views-on-impact/" target="_blank">Pew Research</a>.</li><li>The clearest gains are being captured by investors, tech giants and power users, while ordinary Americans are being asked to absorb the disruption to jobs, energy and information feeds.</li></ul><p><strong>What to watch: </strong>The Trump administration's Labor Department published a <a href="https://www.dol.gov/newsroom/releases/eta/eta20260213" target="_blank">national AI literacy framework</a> in February, aimed at helping workers "share in the prosperity that AI will create."</p><ul><li>OpenAI, Anthropic, Microsoft and Amazon helped pool $500 million in June for <a href="https://www.axios.com/2026/06/25/anthropic-labor-market-ai-jobs-crisis" target="_blank">RAISE US</a>, a workforce retraining initiative led by former Commerce Secretary Gina Raimondo and former Indiana Gov. Eric Holcomb.</li><li>But basic literacy efforts can only go so far: Frontier users have better tools, earlier access, deeper technical context and hundreds of hours of trial-and-error with systems that change every few weeks.</li></ul><p><strong>Flashback: </strong>A century ago, electricity exposed a similar divide between Americans living in the modern age and those watching it from the dark.</p><ul><li>By 1930, nearly 90% of urban homes had electricity, compared with roughly 10% of farms. Private utilities had little incentive to wire rural customers spread across miles of unprofitable territory.</li><li>It took the New Deal's <a href="https://eh.net/encyclopedia/rural-electrification-administration/" target="_blank">Rural Electrification Administration</a> — and years of federal loans — to bridge a gap the market had left behind.</li><li>AI's divide may be even harder to close: Frontier access is scarce and expensive, and even where it's free, most people don't know what to do with it.</li></ul><p><strong>The bottom line: </strong>The AI industry is betting on inevitability. But history suggests technological revolutions need legitimacy, too.</p>
<p>Three <a href="https://www.axios.com/2026/05/25/2028-trend-convergence-ai-politics-platform-shift" target="_blank">AI trends</a> are accelerating and colliding, forcing government, business and investors to rethink strategies in real time:</p><ol><li>AI is getting bigger and better, both here and in China. </li><li>The U.S. government is scrambling to keep pace by creating a regulatory framework, perhaps with international reach.</li><li>Both America and China are considering blocking access to their best AI, in recognition of the rising stakes. </li></ol><p><strong>Why it matters:</strong> The explosive rise of truly autonomous agents is forcing Washington and Beijing away from light-touch oversight, transforming the global AI race from a commercial sprint into a national-security standoff.</p><hr /><p><strong>Here's our latest intel on each trend,</strong> based on conversations with top AI execs and administration sources, and our team's stress-testing of advanced AI models:</p><p><strong>1. Models muscle up:</strong> Increases in the capability of the big AI models (led by OpenAI's ChatGPT, Anthropic's Claude and Google's Gemini) tend to get covered incrementally by the media. But we've just lived through a transformational few months.</p><ul><li>Anthropic's <a href="https://www.axios.com/2026/06/27/anthropic-fable-5-return-soon" target="_blank">Fable</a> and <a href="https://www.axios.com/2026/04/08/mythos-system-card" target="_blank">Mythos</a> models — restricted in June for nearly three weeks over security concerns — have set a new standard for the mind-boggling power of frontier AI. Engineers can hand these models entire multimillion-line codebases and walk away for days, trusting agents to rebuild outdated systems, fix their own bugs and test their own work with shockingly little oversight.</li><li>After a "voluntary" delay due to government consultations, <a href="https://www.axios.com/2026/07/08/openai-gpt-trump-ban-lifted" target="_blank">OpenAI</a> came roaring back with Sol — a model <a href="https://www.axios.com/2026/07/08/gpt-sol-ultra-openai-anthropic-grok" target="_blank">early testers describe</a> as a quantum leap in agentic power. Developers have been left slack-jawed by its ability to summon swarms of sub-agents that collaborate, hunt for security flaws and rewrite software at speeds that make previous models feel like dial-up.</li><li>Elon Musk's SpaceXAI, fresh off its <a href="https://www.axios.com/2026/06/11/spacex-ipo-prices-75-billion" target="_blank">record-breaking IPO</a> and <a href="https://www.axios.com/2026/06/16/spacex-cursor-60-billion-musk" target="_blank">$60 billion acquisition of Cursor</a>, clawed its way back into the AI race Wednesday with the release of <a href="https://www.axios.com/2026/07/08/spacexai-grok-new-model" target="_blank">Grok 4.5</a> — a model triple the size of its predecessor. <a href="https://x.com/elonmusk/status/2074893922812252354" target="_blank">Musk says</a> another model nearly twice as large is coming next month, doubling down on a bet that raw scale, not just smarter training, still wins.</li><li>Meanwhile, China is dominating the open-source race. <a href="https://www.axios.com/2026/06/25/china-glm-52-open-source-hackers" target="_blank">GLM-5.2</a>, built by Chinese startup Z.ai, is free to download and now performs in the same tier as America's priciest models. Z.ai founder Jie Tang <a href="https://x.com/jietang/status/2067580270078030088?s=20" target="_blank">predicted</a> China will achieve a "Fable-class" model before Q1 of 2027.</li></ul><p><strong>2. Administration activating:</strong> President Trump initially took a <em><a href="https://www.whitehouse.gov/releases/2025/07/white-house-unveils-americas-ai-action-plan/" target="_blank">laissez-faire </a></em><a href="https://www.whitehouse.gov/releases/2025/07/white-house-unveils-americas-ai-action-plan/" target="_blank">approach</a> to AI as a way of keeping America's lead over China. But we've learned that top officials are vigorously debating a much more systemic and prescriptive approach, including protocols for the AI labs to follow before releasing their most powerful models. </p><ul><li>"The possibilities are wide open," said an outside adviser deeply involved in the conversations.</li><li>Trump is <a href="https://www.whitehouse.gov/presidential-actions/2025/01/unleashing-prosperity-through-deregulation/" target="_blank">reluctant to regulate</a>, as is clear from <a href="https://www.wsj.com/politics/policy/top-doj-official-tells-staff-he-wants-to-avoid-antitrust-trials-bc5a23ce?st=sQvwQA&reflink=desktopwebshare_permalink" target="_blank">his approach</a> across <a href="https://www.foxbusiness.com/economy/trump-admin-maps-out-sweeping-rollback-regulations-push-save-1-5t" target="_blank">much of the Executive Branch</a>. But the power of Claude Mythos 5 and Fable 5 has roused many officials to favor a more robust, less ad hoc approach.</li><li>Restrictions on those models showed the administration's hand: If national security becomes an issue, complying with the government becomes mandatory, as evidenced by Commerce Secretary Howard Lutnick's letters to Anthropic. A U.S. official told us: "The <a href="https://www.anthropic.com/news/fable-mythos-access" target="_blank">export controls</a> were effective [in] ensuring Anthropic worked with the administration."</li><li>Mythos, we were told again and again, was a wake-up call that more guardrails are needed. We've learned Trump officials are considering a new governing body for vetting AI, with the possibility of including other nations.</li></ul><p><strong>3. U.S., China contemplate controls:</strong> Chinese authorities have met with top tech firms over the past month to discuss restricting overseas access to the country's most advanced AI models, <a href="https://www.reuters.com/world/beijing-is-looking-curbing-overseas-access-chinas-top-ai-models-sources-say-2026-07-07/" target="_blank">Reuters reported</a> this week.</p><ul><li>When we started asking around about the report, we were surprised to hear the U.S. is kicking around ways to restrict Chinese access to U.S. models, perhaps through export controls. These conversations are very preliminary, with little agreement about what measures could actually work.</li><li>This isn't just about American competitiveness — national security is at stake. "AI is already deeply integrated into both countries' intelligence and military, which will change the geopolitical competition and the nature of warfare," said an insider who talks with competing factions of the administration.</li></ul><p><strong>The bottom line: </strong>We've entered the Big Phase — big government considering new rules, big AI in a neck-and-neck race for frontier supremacy, and the big global showdown of China vs. USA.</p><ul><li><em>Axios' Zachary Basu and Andrew Kay contributed reporting.</em></li></ul><p><em><a href="https://www.axios.com/2026/06/18/jim-vandehei-ai-writing-tips-claude-chatgpt" target="_blank">Go deeper</a>: Writing with AI.</em></p>
<p>The fight that scrubbed the world's most powerful AI models from the internet featured personality clashes, industry confusion and international backlash.</p><p><strong>Why it matters: </strong>Anthropic's models are back online, but the impact of its 20-day showdown with the Trump administration will be long lasting.</p><hr /><p><strong>Behind the scenes: </strong>It began when Amazon, Anthropic's partner and investor, sounded an alarm that was later disputed by cybersecurity experts.</p><ul><li>It warned about a "jailbreaking" issue it found with the AI lab's latest models, Mythos and Fable — meaning a technical flaw that could have caused a failure of their guardrails.</li><li>Amazon flagged its concerns to the administration, <a href="https://www.axios.com/2026/06/12/anthropic-trump-mythos-fable-national-security" target="_blank">triggering sweeping export controls</a>. A U.S. official said the government conducted its own tests once it became apparent that the issue needed to be addressed.</li><li>Cybersecurity experts, however, <a href="https://www.axios.com/2026/06/15/anthropic-fable-security-leaders-trump-admin" target="_blank">later wrote</a> in an open letter to the administration that other leading AI models have the same issue Amazon warned about with Anthropic.</li></ul><p><strong>On June 12, </strong>Commerce Secretary Howard Lutnick, at the direction of President Trump, called Anthropic CEO Dario Amodei. </p><ul><li>Lutnick made clear to Amodei the issue needed to be resolved fast and alerted the CEO that the company would be receiving a letter imposing sweeping export controls, the U.S. official said.</li><li>Amodei called Lutnick back that night after receiving the letter, realizing it effectively meant the models would have to be taken offline — to which Lutnick responded that was indeed the goal.</li></ul><p><strong>That decision led to </strong>a three-week, multi-agency crash course in AI safety.</p><ul><li>Anthropic <a href="https://www.axios.com/2026/06/14/anthropic-white-house-mythos-fable" target="_blank">deployed engineers</a> to Washington D.C. According to a U.S. official, the company wanted to prove everything was already resolved and further changes were being fine tuned.</li><li>But the federal Center for AI Standards and Innovation and the National Security Agency said those changes weren't good enough, prompting further fixes, according to the U.S. official.</li><li>Gradually, various agency heads approved of the changes, and on July 1 the models were released, the official said.</li></ul><p><strong>Out of all </strong>of the administration officials Amazon's Andy Jassy could have called,<strong> </strong>it was Treasury Secretary Scott Bessent who first heard about the jailbreaking issue found in the company report, according to a separate source familiar.</p><ul><li>Bessent was early to sound the alarm on<strong> </strong>Mythos, work with White House chief of staff Susie Wiles to <a href="https://www.axios.com/2026/04/17/anthropic-white-house-wiles-bessent-amodei" target="_blank">reengage the embattled company</a> and help get a <a href="https://www.axios.com/2026/06/02/trump-signs-new-ai-executive-order" target="_blank">cybersecurity executive order</a> across the finish line.</li><li>While technical discussions to address the jailbreaking issue took place in D.C., it was Bessent who stood next to Trump during the G7, where allies called for global cooperation on safety standards.</li></ul><p><strong>At the center of the showdown was </strong><a href="https://www.axios.com/2026/06/18/inside-white-house-ai-power-center" target="_blank">Lutnick</a>, who also flanked Trump at the G7 meeting while his department's teams led technical discussions.</p><ul><li>National cyber director Sean Cairncross, the White House Office of Science and Technology Policy, Treasury Department chief information officer Sam Corcos and the NSA also all participated in technical discussions, according to various sources.</li><li>Washington mobilized faster to hold scores of meetings and pulled in far more agencies than one would expect for a single technical issue, one source said.</li></ul><p><strong>The tension spiraled </strong>amid <a href="https://www.axios.com/2026/06/15/anthropic-white-house-fable-mythos" target="_blank">personality clashes</a> and poor communication.</p><ul><li>Anthropic eventually understood that in order to be successful, it needed to be on the same side as the government, the U.S. official said.</li><li>As discussions turned more technical, Anthropic policy chief Sarah Heck and<strong> </strong>co-founder Tom Brown got more involved. Brown also had multiple conversations with Lutnick and Cairncross the weekend of June 12.</li><li>There was never a moment when Dario stepped offstage and someone else replaced him, one source said, adding that Brown's technical expertise allowed him to sit in a room with government specialists and go line by line through how models behave under stress.</li></ul><p><strong>Between the lines: </strong>It remains uncertain when and how Anthropic's models will be released to ally countries around the world — which proponents say is key to beating China — or how other labs from OpenAI to Google will release their latest models.</p><ul><li>OpenAI, whose latest model GPT-5.6 is on hold, did not have visibility into discussions between Anthropic and the White House and is engaged in daily technical discussions on the release of its own model, a source said.</li></ul><p><strong>The bottom line: </strong>There's a lot of work left to be done on a framework for approving future models with a clear inclusive process that has transparency standards and timelines, sources familiar said.</p>
<p><a href="https://www.axios.com/2026/06/26/openai-gpt-sol-terra-luna-trump" target="_blank">OpenAI</a> may give the U.S. government a 5% stake in the company, per the <a href="https://www.ft.com/content/7c803eab-8e80-4431-9a87-e943bf00e00b?accessToken=zwAAAZ8jH94Hkc98gD6rjoBEMdOah-lDvwDgCw.MEQCIDs8Drq2784t9EO7dNO60tKqNy-axAWfV0SYDxQtxlUPAiBhUdyW47NO2oXPJ-ifp6vV4YJNWo7wYlOSuwYdcYClPQ&sharetype=gift&token=d72303a2-2849-404f-9115-d89370c60588&syn-25a6b1a6=1" target="_blank">Financial Times</a><strong>. </strong>The proposal is in very preliminary conversations, according to a person familiar with the matter.</p><p><strong>Why it matters: </strong>If the overture is taken up by the <a href="https://www.axios.com/politics-policy/donald-trump" target="_blank">Trump</a> administration, that would mean the government would have a vested interest in weighing whether or not to limit the release of an OpenAI model.</p><hr /><p> <strong>Catch up quick:</strong> OpenAI views the potential government stake as a way to give the general public a share of the upside of AI, and CEO Sam Altman has previously shared <a href="https://www.youtube.com/watch?v=B21KxGs8zDI&vl=en" target="_blank">with Axios</a> his interest in some sort of public wealth fund.</p><ul><li>The goal is to include other <a href="https://www.axios.com/technology/automation-and-ai" target="_blank">AI</a> labs giving over a similar stake.</li><li>This could look like including shares in <a href="https://www.axios.com/2026/07/02/trump-accounts-savings-etf" target="_blank">Trump accounts</a> or some other vehicle that would give American households exposure to investments in AI, for example. </li><li>Anthropic supported similar policies in a recent <a href="https://www-cdn.anthropic.com/files/4zrzovbb/website/9ea607a5dd67c168093829b701f3a0a6d21156d5.pdf" target="_blank">paper</a>, arguing for "universal pre-distributive capital accounts" with "priority given" to those with jobs exposed to AI disruption.<strong> </strong></li></ul><p><strong>Zoom out: </strong>The potential investment comes as the White House is still deciding when OpenAI can release its most powerful models widely, through a regulatory process that Altman has <a href="https://x.com/sama/status/2070607488274358364" target="_blank">said</a> isn't quite "optimal." </p><ul><li>Altman <a href="https://www.ft.com/content/0c2e1077-f658-4b3d-9040-602615c961ca?accessToken=zwAAAZ8jqdiQkc8MLhB39lhLPdOQQGAmFclhyg.MEUCIQDTUiIL0UYEaoE25S4euqGkhgOMxJn4r27SftiGoI39NQIgM8b4a3IiQ8df8aq_EQ7mIljhjCUr7FFPC0kYZkxEgcU&sharetype=gift&token=2b1a04f5-7679-4c10-a83a-d4de72e18f47" target="_blank">proposed</a> on Wednesday a U.S.-led international forum to establish AI regulatory standards, which could be a way to allow the government to invest without having as heavy a hand in regulation.</li></ul><p><strong>Between the lines: </strong>Investors tell Axios that the idea of giving a stake reads like a PR stunt aimed at making it seem as if the public could benefit from the AI boom just as the technology threatens their jobs.</p><p><strong>Yes, but: </strong>The government's <a href="https://www.axios.com/2025/08/14/trump-intel-stake-semiconductor" target="_blank">9.9% stake in Intel</a>, taken last August, appears to have paid off.</p><ul><li>Its shares are up nearly 400% since then, although it has also come amid a broad rally for chip stocks. </li></ul><p><strong>Friction point: </strong>The Intel stake was acquired under the CHIPS Act. A deal with the AI labs would likely require an act of Congress.</p><ul><li>It's also unclear what a government stake would accomplish, other than giving the AI labs a closer relationship to what is currently one of their biggest hurdles: the government. </li><li> An investor in Anthropic and OpenAI tells Axios that the proposal reads more like a "political move" to gain favor with the administration than something that would actually create a shared benefit for the American public. </li></ul><p><strong>What they're saying: </strong>A government stake would be a "troubling milestone" that hurts competition between the labs, " David Sherman, AI and financial inclusion strategist at <a href="https://urldefense.com/v3/__http%3A//io.net__;!!Al82Z4c!0iM9era3UzqeHAPC8khMxFVMB3laKcMhSbvnKhzN94Xjxl5L4EEHRd0sWYiSyUrcKTIPV8tqQVDKqrZlg5lNtd0$" target="_blank">io.net</a>, a decentralized cloud network, said via email.</p><ul><li>It "gives one AI company a government stamp of approval whilst millions of developers, researchers and businesses are locked out by skyrocketing token prices and endless GPU queues," he added, referencing the challenge in accessing AI at current costs or chips at current supply levels.</li><li>It comes amid broader concern about whether the government is already curbing the competitive edge of U.S. AI labs compared with China by putting guardrails around model release timelines.</li></ul><p><strong>And if the goal is to share</strong> the financial benefits of AI, there are other ways to do that. </p><ul><li>Some have suggested that AI companies share a percent of <a href="https://x.com/thestalwart/status/2072649035027267972?s=46&t=lWeRdBU5epOoS27GJ8FHHg" target="_blank">pre-tax income</a> or that the government impose a tax on all tokens. Kevin Bankston, an AI governance advisor, wrote "JUST. TAX. THEM." on <a href="https://x.com/KevinBankston/status/2072660701965938973" target="_blank">X</a>.</li><li>Bill Gates proposed an automated <a href="https://www.weforum.org/stories/2017/02/bill-gates-this-is-why-we-should-tax-robots/" target="_blank">robot tax </a>in 2017 that could slow down automation, suggesting they should be taxed the same way human employees pay income tax.</li></ul><p> <strong>The bottom line: "</strong>The labs develop the technology, but citizens and their elected representatives must make the rules," Altman wrote in the Financial Times earlier this week. </p>
<p>President Trump is redefining what it means to be a U.S. ally in the <a href="https://www.axios.com/technology/automation-and-ai" target="_blank">AI</a> era.</p><p><strong>Why it matters: </strong>For the <a href="https://www.axios.com/2026/06/18/inside-white-house-ai-power-center" target="_blank">White House</a>, it's now about how <a href="https://www.axios.com/2026/06/26/china-ai-us-alliance" target="_blank">partners</a> can help the U.S. win the AI race.</p><hr /><ul><li>For decades, shared values and security interests have underpinned alliances with <a href="https://www.axios.com/2026/06/12/5-takeaways-europe-tech-chief-henna-virkkunen-ai" target="_blank">Europe</a> and other partners around the world. Under Trump, that's no longer enough.</li><li>As AI becomes central to economic and military power, frontier AI models, chips and infrastructure are turning into new instruments of American influence.</li></ul><p><strong>Driving the news: </strong>The Trump administration is <a href="https://www.axios.com/2026/06/15/anthropic-white-house-fable-mythos" target="_blank">blocking</a> allies from accessing the world's most powerful models, playing it close to the vest and criticizing Europe for not having its own robust AI industry.</p><ul><li>With export controls on Fable and Mythos lifted on Tuesday, Anthropic and the Trump administration are continuing <a href="https://www.anthropic.com/news/expanding-project-glasswing" target="_blank">Project Glasswing</a> efforts, which the company previously said would give access to Mythos to 150 more organizations across more than 15 countries.</li><li>Fable is also <a href="https://www.axios.com/2026/06/30/trump-anthropic-ai-model-fable-restrictions" target="_blank">back</a>.</li><li>Commerce retains the power to pull back access when it deems appropriate. </li></ul><p><strong>"The problem we have</strong> is that we are leading everybody by a lot," Trump said in a recent <a href="https://www.axios.com/2026/06/19/trump-axios-show-interview-transcript-marc-caputo" target="_blank">interview</a> with "The Axios Show." "Europe has to be very careful. They're losing their way entrepreneurially."</p><ul><li>Trump pointed to the U.K. not tapping into energy sources in the North Sea because of environmental concerns: "It's crazy." </li></ul><p><strong>Between the lines:</strong> The Trump administration's AI restrictions are part of a broader transactional approach to alliances.</p><p><strong>Catch up quick: </strong>Vice President JD Vance's <a href="https://www.presidency.ucsb.edu/documents/remarks-the-vice-president-the-artificial-intelligence-action-summit-paris-france" target="_blank">speech</a> last year at the Paris AI Summit set the stage for a confrontational relationship with the European Union.</p><ul><li>The U.S. was quick to rebuke the EU's focus on safety over innovation with Trump entering his second term laser-focused on deregulation. </li><li>But the president now finds himself behind an ad hoc licensing regime that's creating its own regulatory uncertainty, both domestically and abroad.</li><li>OpenAI's GPT-5.6 was forced into a staggered <a href="https://www.axios.com/2026/06/25/trump-administration-openai-gpt-model-release" target="_blank">rollout</a> last month due to government concerns.</li></ul><p><strong>What they're saying: </strong>European Commission spokesperson Thomas Regnier on Wednesday said the bloc is looking forward to "intensifying" discussions with Anthropic to gain access to Mythos through Project Glasswing and that, in the meantime, it has access to GPT-5.5-Cyber.</p><ul><li>The EU has a delegation in the U.S. to determine the scope of a future tech dialogue, the frequency of meetings and the level at which they'll be held.</li><li>The dialogue is expected to include frontier AI models, chip supply chains and cybersecurity.</li><li>"But we have one clear line," Regnier said, "which is that our sovereign legislation is not up for negotiation."</li></ul><p><strong>Omran Sharaf, the United Arab Emirates' assistant foreign minister</strong> for advanced science and technology, told Axios "it's very important that trusted partners and strategic partners are included in the process."</p><ul><li>That way it is "synchronized and we're applying similar standards in controlling such technologies, rather than having something that gets imposed."</li></ul><p><strong>The big picture: </strong>AI is changing what the White House wants in its alliances.</p><ul><li>Just last week, the EU and several European governments signed onto Pax Silica, the U.S.-led effort to secure AI supply chains and critical minerals, even as the White House restricts their access to frontier AI models.</li></ul><p><strong>For the U.S., Europe is all of the above:</strong> a restraint, an indispensable partner and a competitor.</p><ul><li>The administration is simultaneously rejecting Europe's AI rules, inking deals with the region to secure supply chains and blocking access to cutting-edge technology.</li></ul><p><strong>The bottom line:</strong> With the most advanced AI, allies will have to adjust to being considered trusted U.S. partners only in some cases.</p><ul><li>"The conclusion that governments are coming to is: We'll be part of Pax Silica, yes. We'll work with the U.S. ecosystem where we can. We'll build around what we can't," A.J. Bhadelia, Cohere's head of global government affairs and external affairs, told Axios.</li></ul>
<p>Anthropic's Fable 5 model came back online for users on Wednesday, after the Trump administration <a href="https://www.axios.com/2026/06/30/trump-anthropic-ai-model-fable-restrictions" target="_blank">lifted an export control</a> late Tuesday.</p><p><strong>Why it matters: </strong>It's the most powerful publicly available <a href="https://www.axios.com/technology/automation-and-ai" target="_blank">AI</a> tool — so capable that the U.S. government decided Anthropic had to add further safety measures in order to make it broadly available.</p><hr /><p><strong>Driving the news:</strong> <a href="https://www.axios.com/2026/06/27/anthropic-fable-5-return-soon" target="_blank">Fable</a> is available to all customers, Anthropic said, though queries it deems to pose security or safety risks may be routed to less powerful models.</p><ul><li>For the most part, customers who want to use Fable will have to do so outside of any subscription plan, paying for the tokens they use.</li><li>Anthropic says that, until July 7, subscribers can tap Fable for up to half of their included data usage, though it warns Fable will burn through tokens faster than other models.</li></ul><p><strong>The big picture:</strong> While Fable is once again available, its blockage by the U.S. government has raised significant questions over how and when the Trump Administration will step in to block future frontier model releases.</p><ul><li>OpenAI has said it is withholding broad release of its latest model, <a href="https://www.axios.com/2026/06/26/openai-gpt-sol-terra-luna-trump" target="_blank">GPT 5.6</a>, while it consults with the government, which <a href="https://www.axios.com/2026/06/25/trump-administration-openai-gpt-model-release" target="_blank">requested</a> the delay.</li></ul>
The U.S. government slapped Anthropic with export restrictions days after the Fable 5 model was released to the greater public over cybersecurity safety concerns.