I spent much of the past week in Washington, D.C., where, when it comes to AI, change is in the air. The Trump administration is in the process of pivoting from an AI policy built largely around opposing and dismantling AI regulation, to a possible federal licensing regime for AI models. Meanwhile, support for AI regulation is building on both sides of the aisle on Capitol Hill, and if the Democrats seize at least one Congressional house in November, the passage of AI legislation of some kind is almost guaranteed. Not only that, but President Donald Trump\u2019s visit to China last week seems to have signaled a significant shift in the administration\u2019s thinking on international AI governance too.<\/p>\n
This transition is driven by two things. One is the public backlash against AI, which has gained significant momentum in the past few months, as a story in the Wall Street Journal<\/em> earlier this week chronicled. The viral video of former Google CEO Eric Schmidt\u2019s commencement address at the University of Arizona in which the graduating students roundly booed him every time he mentioned AI is just one data point. A litany of recent polls have shown that most Americans are more fearful than hopeful about AI, with the gap sometimes as wide as 40 percentage points. Seven in 10 Americans oppose the construction of data centers in their local community. An Annenberg Public Policy Center poll released last week found that two-thirds of Americans thought the government had done too little to regulate AI, a view held by the majority of Republicans and independents, as well as Democrats.<\/p>\n Politicians know they can\u2019t afford to be on the wrong side of numbers like this. Trump, who has populist instincts, is starting to see this too. It\u2019s one of the reasons that the more populist-minded and politically savvy of Trump\u2019s advisors, people like Treasury Secretary Scott Bessent and Susie Wiles, have moved to wrestle AI policy away from tech bro advisors such as David Sacks, Sriram Krishnan, and Michael Kratsios. Denizens of Silicon Valley\u2019s venture capital scene, they simply don\u2019t have the instincts for how AI is playing out on Main Street and how it could, if the White House gets it wrong, pose a major problem for the GOP in November and beyond.<\/p>\n You can also see this effect in OpenAI\u2019s decision last week to endorse both the bipartisan Kids Online Safety Act (KOSA), which would require all online platforms that potentially have children as users to take steps to prevent and mitigate any harm to them, as well as a state AI bill, SB 315, currently pending before the Illinois state legislature. That bill would require companies building frontier AI models to establish safety frameworks, conduct annual audits, report any critical incidents, and protect whistleblowers. Many tech industry associations have been lobbying against KOSA, saying its tenets are vague and potentially unconstitutional. Meanwhile, OpenAI had previously opposed safety legislation in California that was substantially similar to the Illinois bill. But apparently OpenAI has belatedly figured out which way the wind is blowing.\u00a0<\/p>\n The second thing that has changed is Mythos. Anthropic\u2019s powerful AI model, with its superhuman hacking skills, has woken the government up to the fact that AI is a potent dual use technology and that it cannot leave decisions about when, how, and where it gets released totally up to the tech companies creating it. The Trump administration has already reportedly vetoed Anthropic\u2019s plans to expand \u201cProject Glasswing,\u201d a program under which it shared a\u00a0version of Mythos with select companies to help them find and patch software vulnerabilities, possibly out of concern that model would be more likely to fall into the wrong hands (and possibly to preserve the National Security Agency\u2019s offensive cyber capabilities.) And Bessent is playing a big role in deciding which foreign financial authorities and banks are getting access to the model. In essence, Mythos is already being subjected to an ad hoc licensing regime.<\/p>\n Brad Carson, the former Oklahoma Congressman who now heads Americans for Responsible Innovation, a group that advocates for tech regulation, told me he thinks Mythos is the \u201cEl Alamein moment\u201d in the fight for AI regulation. El Alamein is the World War II battle, which took place in the fall of 1942, in which British forces first proved that they were capable of defeating the Germans. Churchill called the battle \u201cthe end of the beginning\u201d and noted that before El Alamein, the British had never had a victory, but that afterwards, they never had a defeat. Carson says, of the battle over AI regulation, \u201cit\u2019s not over yet, the way that El Alamein was not the end of the war. But the fall of Berlin is in sight, and [regulation of AI] is going to happen.\u201d<\/p>\n The Mythos effect is not confined to domestic policy. Mythos has shifted the prospects for international AI governance as well. One of the most interesting developments to come out of Trump\u2019s summit with Chinese President Xi Jinping last week was that the U.S. apparently agreed\u2014according to the New York Times<\/em>\u2014to hold talks on AI safety with Beijing. Before, the accelerationist crew in command of Trump\u2019s AI policy were firmly opposed to any discussions with China, believing any treaty would only serve to hobble U.S. AI efforts, while China would likely renege on any promises it made. They also liked to use the \u201cChina card\u201d as a reason for opposing any domestic AI regulation.<\/p>\n But Carson thinks the China card has lost its salience. Most Americans are more afraid of AI job losses than they are of China getting ahead in AI. Meanwhile, Mythos seems to have convinced both Chinese and American officials that it is in neither of their interests for non-state actors to get a hold of dangerous cyber capabilities.<\/p>\n It\u2019s also perhaps dawning on some people in the administration that U.S. AI policy with regards to China isn\u2019t working as intended. Both the Trump administration and the Biden administration have used export controls on AI chips and on chipmaking equipment ostensibly to prevent China from developing powerful AI systems that might give them a military advantage over the U.S.<\/p>\n \u201cBut there is little evidence that the export controls have actually delayed or made it more difficult for China to acquire militarily useful AI capabilities,\u201d says Jacob Feldgoise, a senior data analyst at Georgetown University\u2019s Center for Security and Emerging Technology. Many of the AI systems used in military applications, such as autonomous navigation for drones, or analyzing satellite imagery to find targets, don\u2019t depend on the kinds of large language models that require large volumes of advanced GPUs to train and run. And when it does come to decision support systems run by LLMs, Chinese tech companies are only about six months behind America\u2019s AI labs in developing frontier capabilities. That may give the U.S. some ability to potentially relax export controls in exchange for Chinese cooperation on establishing an international governance regime for AI. But exactly what that framework may look like is still an open question.<\/p>\n What is clear is that the mood has shifted dramatically, and Washington\u2019s hostility to AI regulation has crumbled.<\/p>\n Ok, with that, here\u2019s this week\u2019s AI news.<\/p>\n Jeremy Kahn<\/strong> Clarification, May 27: <\/strong>This story has been updated to clarify that Jacob Felgoise was suggesting that export controls were intended to delay China from obtaining military AI capabilities, not prevent it from acquiring strategically useful ones. Also the text has been modified to make it more clear when Feldgoise is being quoted directly and when the author is paraphrasing or adding explanation that does not reflect Feldgoise\u2019s exact words. <\/em><\/p>\n Before we get to the news, a quick note: Join us on Thursday, May 28<\/strong>, for \u201cFortune 500 Europe: In Conversation with Tech Leaders,\u201d a candid virtual exchange with senior technology leaders from Fortune 500 Europe companies, including Mars Pet Nutrition, Orange, Reckitt, and Saint-Gobain. The discussion will explore one of the most pressing questions organizations face today: how to turn AI investment into sustainable business value. Register your interest to attend and receive Fortune<\/em>\u2018s editorial takeaways.<\/p>\n Jury rules against Elon Musk in $150 billion lawsuit against OpenAI and Sam Altman<\/span>\u2014by Sharon Goldman<\/span><\/p>\n Exclusive: AI startup Viktor raises $75 million to put a virtual \u2018coworker\u2019 in Slack and Teams<\/span>\u2014by Beatrice Nolan<\/span><\/p>\n Parag Agrawal\u2019s AI startup wants to pay publishers when AI agents use their work<\/span>\u2014by Beatrice Nolan<\/span> Google and Blackstone strike $5 billion deal to build new cloud company. <\/b>Google is partnering with asset management and private equity firm Blackstone to launch a new U.S.-based AI cloud company that will aim to compete with other so-called \u201cneocloud\u201d providers such as CoreWeave and Crusoe, the <\/span>Wall Street Journal<\/span><\/i> reports<\/span>. Blackstone is putting $5 billion in equity into the new venture and will be the majority owner. The new company will use Google\u2019s own AI chips, called tensor processing units, or TPUs, along with Google software and services. The company expects to bring 500 megawatts of capacity online by 2027 and scale toward tens of billions of dollars in total AI infrastructure investment. The new venture is the second big investment from BXN1, a new Blackstone unit that is overseeing the firm\u2019s AI bets; it earlier formed a joint venture with Anthropic to help sell AI tools to companies.<\/span> OpenAI reorganizes its product teams ahead of likely IPO. <\/b>OpenAI has put cofounder Greg Brockman in charge of overall product strategy as it aims to unify its offerings around an \u201cagentic\u201d AI future. The company is merging ChatGPT, Codex, and its developer API into a single core product organization, reflecting increasing overlap between consumer and enterprise tools. The shake-up also includes other leadership changes: Thibault Sottiaux, who had been leading Codex, will now lead core product and platform teams; Nick Turley, the former ChatGPT head, will now lead OpenAI\u2019s enterprise products; Ashley Alexander, who had led health products, will now take the reins of the consumer product team. The move comes amid rising competition from rivals like Anthropic and Google and ahead of a potential IPO later this year. Read more from <\/span>Wired<\/span><\/i> here<\/span>.<\/span> AI21 Labs slashes staff, pivots to agents. <\/b>The Israeli AI company, which had been building its own language models, is cutting more than 60% of its workforce\u2014reducing staff from about 180 to 70\u2014as part of a major restructuring and strategic pivot toward optimizing AI agents that may use third-party AI models. The shift follows the collapse of acquisition talks with Nebius, though the two firms have agreed to a commercial partnership. AI21 will stop selling standalone language models and focus on optimizing enterprise AI agents, betting this approach offers a more sustainable business model. Read more in trade publication CTech <\/span>here<\/span>.\u00a0<\/span><\/p>\n xAI promised to pay employees for tax data, but hasn\u2019t. <\/b>That\u2019s according to <\/span>a scoop<\/span> from Bloomberg. xAI asked employees to submit their tax returns as training data for its Grok chatbot, promising a $420 payment and perks, but two months later those payments have not been made. The initiative was part of a push to improve Grok\u2019s tax capabilities and compete with rivals like Anthropic and OpenAI, with the request later extended to employees\u2019 friends and family. The missing payments have hurt morale inside xAI, which is already dealing with layoffs, management turnover, and a broader restructuring effort.<\/span><\/p>\n Analog Devices in talks to acquire AI chip company Empower. <\/b>That\u2019s according to an exclusive <\/span>story<\/span> in The Information that says Analog Devices is in advanced talks to acquire Empower Semiconductor for about $1.5 billion. Empower\u2019s chips improve energy efficiency by delivering power directly within or beneath AI processors, reducing losses and stabilizing performance during heavy workloads. The deal reflects surging demand for solutions that handle the massive energy requirements of AI systems and would help Analog compete with rivals like Monolithic Power Systems.<\/span><\/p>\n Google launches its Co-Scientist tool and reports early successes. <\/b>Google DeepMind today unveiled Co-Scientist, a multi-agent AI system built on its Gemini models that’s designed to help scientists generate, refine, and test new research hypotheses. Detailed in a <\/span>Nature<\/span><\/i> paper, also published today, the tool deploys a coalition of specialized agents: one proposes ideas, another acts as a virtual peer reviewer, and a third runs what DeepMind calls a “tournament of ideas” inspired by its game-playing AlphaGo and AlphaStar systems. The system can tap web search and biomedical databases such as ChEMBL and UniProt, and even call on AlphaFold for protein-structure predictions.<\/span><\/p>\n Early results are striking. At Stanford, medical researcher Gary Peltz used Co-Scientist to identify an existing drug that could be repurposed to help treat liver fibrosis. In lab tests, the drug Co-Scientist blocked 91% of the responses that lead to liver scarring. Meanwhile, Calico Life Sciences, the University of Edinburgh, and the University of Cambridge have reported similar wins, including a novel hypothesis about the cellular stress response that was later confirmed in the lab.<\/p>\n Google is making the tool available to individual researchers through Gemini for Science and previewing an enterprise version with Daiichi Sankyo, Bayer Crop Science, and the U.S. National Laboratories’ Genesis Mission. You can read in Google DeepMind’s blog on Co-Scientist here.\u00a0<\/span><\/b><\/p>\n June 8-10<\/strong>: Fortune Brainstorm Tech, Aspen, Colo. Apply to attend here.<\/p>\n June 17-20:<\/strong> VivaTech, Paris.<\/p>\n July 6-11:\u00a0<\/strong>International Conference on Machine Learning (ICML), Seoul, South Korea.<\/p>\n July 7-10:\u00a0<\/strong>AI for Good Summit, Geneva, Switzerland.<\/p>\n Aug. 4-6:\u00a0<\/strong>Ai4 2026, Las Vegas.<\/p>\n Is interpretable AI crucial for AI safety, or a blocker to \u201csuperintelligence\u201d? <\/b> This is, in essence, the problem presented by AlphaGo\u2019s famous Move 37. To human Go experts, the move looked like a mistake. It turned out to be brilliant and a key to AlphaGo\u2019s eventual victory. I have written about this dilemma in the context of using AI in business for <\/span>Fortune<\/span><\/i> before. See <\/span>this story<\/span> from 2019.<\/span><\/p>\n The debate kicked off by Roon\u2019s post seemed to miss a few things. As some pointed out, if a system is that smart, it ought to be able to explain its reasoning. The counter to this is that human experts often do things instinctually and cannot always explain why they are doing them\u2014they just know they feel like the right thing to do in that situation. But, at the very least, the AI ought to be able to provide humans with some kind of measure of how confident the AI is in its own decisions. (Google DeepMind used a confidence metric to help human biologists using its protein folding AI AlphaFold get a sense for when to trust the system and when to be more skeptical.)<\/span><\/p>\n The more interesting question might be around alignment\u2014teaching AI systems to follow human values and not to act against human interests. Roon is correct that a kind of naive alignment that says the system should never override human instruction might in some cases produce sub-optimal results. But the idea of a system that tries to achieve some \u201cgreater good\u201d by ignoring what humans say also seems like a fraught solution. (Mama knows best works for children. But are we willing to allow all of humanity to be infantilized in our pursuit of greater knowledge or more optimal solutions?)\u00a0<\/span><\/p>\n From global corporations to local entrepreneurs, artificial intelligence is changing the way businesses operate, compete, and succeed. Explore all of <\/span>Fortune AIQ<\/b>, and read the latest collection of stories below:<\/span><\/p>\n \u2013After AI stole his clients, one Big Tech ghostwriter is using AI to get them back<\/strong><\/span><\/p>\n \u2013Outnumbered: At $4 billion ClickUp, a 3:1 agent-to-human ratio is rewiring work itself<\/strong><\/span><\/p>\n \u2013How a mom-and-pop car wash chain went from sticky notes to AI-powered operations that are upleveling every part of the company<\/strong><\/span><\/p>\n \u2013Solo founders are using AI to do the work of entire teams\u2014but going it alone has limits<\/strong><\/span><\/p>\n \u2013How EarthRanger uses AI to help protect endangered species\u2014and boost the wildlife tourism industry<\/strong><\/p>\n \u2013The smartphone\u2019s days are numbered. Meet the device that could come next<\/strong> #backlash #growing #politicians #White #House #Capitol #pivot #regulation<\/p>\n","protected":false},"excerpt":{"rendered":" Hello and welcome to Eye on AI. In this edition\u2026momentum builds for U.S. AI regulation\u2026Musk loses his lawsuit against OpenAI\u2026Andrej Karpathy goes to Anthropic\u2026Google debuts a \u201cCo-Scientist\u201d\u2026and are humans the… <\/p>\n","protected":false},"author":1,"featured_media":4294,"comment_status":"open","ping_status":"open","sticky":false,"template":"","format":"standard","meta":{"footnotes":""},"categories":[2],"tags":[4860,6668,743,784,872,1338,735,2347,6667,1797,748,2952,1337],"class_list":["post-4293","post","type-post","status-publish","format-standard","has-post-thumbnail","hentry","category-finance-news","tag-backlash","tag-capitol","tag-elon-musk","tag-eye-on-ai","tag-growing","tag-house","tag-openai","tag-pivot","tag-politicians","tag-regulation","tag-sam-altman","tag-trial","tag-white"],"_links":{"self":[{"href":"https:\/\/www.fintechpulse8.com\/index.php?rest_route=\/wp\/v2\/posts\/4293","targetHints":{"allow":["GET"]}}],"collection":[{"href":"https:\/\/www.fintechpulse8.com\/index.php?rest_route=\/wp\/v2\/posts"}],"about":[{"href":"https:\/\/www.fintechpulse8.com\/index.php?rest_route=\/wp\/v2\/types\/post"}],"author":[{"embeddable":true,"href":"https:\/\/www.fintechpulse8.com\/index.php?rest_route=\/wp\/v2\/users\/1"}],"replies":[{"embeddable":true,"href":"https:\/\/www.fintechpulse8.com\/index.php?rest_route=%2Fwp%2Fv2%2Fcomments&post=4293"}],"version-history":[{"count":0,"href":"https:\/\/www.fintechpulse8.com\/index.php?rest_route=\/wp\/v2\/posts\/4293\/revisions"}],"wp:featuredmedia":[{"embeddable":true,"href":"https:\/\/www.fintechpulse8.com\/index.php?rest_route=\/wp\/v2\/media\/4294"}],"wp:attachment":[{"href":"https:\/\/www.fintechpulse8.com\/index.php?rest_route=%2Fwp%2Fv2%2Fmedia&parent=4293"}],"wp:term":[{"taxonomy":"category","embeddable":true,"href":"https:\/\/www.fintechpulse8.com\/index.php?rest_route=%2Fwp%2Fv2%2Fcategories&post=4293"},{"taxonomy":"post_tag","embeddable":true,"href":"https:\/\/www.fintechpulse8.com\/index.php?rest_route=%2Fwp%2Fv2%2Ftags&post=4293"}],"curies":[{"name":"wp","href":"https:\/\/api.w.org\/{rel}","templated":true}]}}Mythos: AI\u2019s \u2018El Alamein\u2019 moment<\/strong><\/h2>\n
A policy that doesn\u2019t hit its stated target<\/strong><\/h2>\n
\u201cBut while there isn\u2019t much evidence that American export controls have prevented China from developing military AI applications, it is likely that export controls do slow the deployment of AI commercially across the economy because China lacks enough GPUs for widespread inference,\u201d Feldgoise says.\u00a0<\/p>\n
jeremy.kahn@fortune.com
@jeremyakahn<\/p>\nFORTUNE ON AI<\/h3>\n
<\/span><\/p>\nAI IN THE NEWS<\/h3>\n
<\/span>
<\/span>Legendary AI researcher Andrej Karpathy joins Anthropic. <\/b>Karpathy, who taught at Stanford, was one of the founding team at OpenAI, headed AI at Tesla for a time, and has a cult following among AI developers, is joining Anthropic, he <\/span>said on X<\/span> and the company confirmed Monday. Karpathy has been working on his own projects for the past several years, including producing educational content on AI.<\/span> At Anthropic, he will be starting a team focused on using Anthropic\u2019s AI model Claude to accelerate pretraining research, the company said in a statement.<\/span><\/p>\n
<\/span>
<\/span>Meta moves staff to Applied AI roles amid pending layoffs. <\/b>Meta is planning to shift more than 7,000 employees into new initiatives, including its Applied AI Engineering (AAI) division and other AI-focused teams, Reuters <\/span>reported<\/span>, citing a memo Meta Chief People Officer Janelle Gale sent to staff earlier this week that the news agency said it obtained. The reassignments and reorganization comes as the company is also preparing to cut about 10% of its roughly 78,000 employees as part of a broader restructuring to align the company more closely with AI-driven priorities. The restructuring and layoffs are expected to be announced on Wednesday.<\/span><\/p>\nEYE ON AI RESEARCH<\/h3>\n
AI CALENDAR<\/h3>\n
BRAIN FOOD<\/h3>\n
<\/b>That was the debate kicked off by \u201cRoon,\u201d the X handle of a member of OpenAI\u2019s technical staff (widely believed to be Tarun Gogineni). <\/span>\u201call else equal, companies and organizations that hand more of themselves over to machine intelligence will outcompete ones that demand the corrigibility and legibility tax of human oversight and human design,\u201d <\/span><\/i>he <\/span>wrote<\/span>.<\/span><\/p>\nFortune AIQ Special Digital Issue: The AI Economy<\/h3>\n
![\"\"](\"https:\/\/fortune.com\/img-assets\/wp-content\/uploads\/2026\/05\/Fortune-AIQ2-Cover.jpg?format=webp&w=1440&q=100\"\/)
<\/div>\n<\/p>\n
<\/span><\/p>\n<\/div>\n