| AI Policy Institute | theaipi.org |
| Your day in AI policy starts here. | Washington, D.C. |
| Bottom Line Up Front |
OpenAI's chatbot dispensed weapon acquisition advice and role-played mass shootings with users reported WSJ, building on prior reporting and a seven-family lawsuit. Anthropic is finalizing a $1.5 billion joint venture with Blackstone, Hellman & Friedman and Goldman Sachs, with Anthropic, Blackstone and Hellman & Friedman each contributing roughly $300 million. Bipartisan AI worry is bridging left and right in Washington, the New York Times reported, with 51% of voters reporting concern about AI.
| Were you forwarded this email? Subscribe → |
Worry over AI is one of the few issues uniting elements of the left and the right in a polarized age, the New York Times reported in a politics desk piece. A Pew poll from last year found 50 percent of Republicans and 51 percent of Democrats were more concerned than excited about the increased use of A.I. in daily life, with just 10 percent reporting more excitement than concern.
Anthropic is launching a roughly $1.5 billion joint venture with Blackstone, Goldman Sachs, Hellman & Friedman and other Wall Street firms to sell AI tools to businesses, particularly private-equity-backed portfolio companies, the Wall Street Journal reported. Anthropic, Blackstone and Hellman & Friedman are each expected to invest around $300 million, with Goldman Sachs putting in roughly $150 million and General Atlantic, Leonard Green, Apollo Global Management, GIC and Sequoia Capital also participating; the venture will operate as a consulting arm helping companies adopt AI across operations. The deal comes as OpenAI pursues a rival joint venture with private-equity firms, with both companies targeting the enterprise market where Anthropic is widely seen as the leader, and as Anthropic eyes a potential public listing as soon as this year on the back of fast-growing Claude Code revenue.
OpenAI is facing growing legal and regulatory scrutiny over ChatGPT's role in mass shootings, with internal staff arguing the company should refer more users to law enforcement, the Wall Street Journal reported. Florida AG James Uthmeier opened a criminal investigation in April after FSU shooter Phoenix Ikner consulted ChatGPT minutes before his April 2025 attack, and seven wrongful-death lawsuits were filed last week in U.S. District Court for Northern California over the February 2026 Tumbler Ridge, B.C. shooting by Jesse Van Rootselaar. OpenAI's investigations team has reportedly pushed internally for more referrals than the roughly 15-30 users the company refers to authorities annually, while its legal team and CEO Sam Altman have favored user privacy. A Center for Countering Digital Hate and CNN study cited in the piece found Anthropic's Claude and Snap's chatbot were the only major systems to reliably refuse violent-attack planning queries, with CCDH's Imran Ahmed saying such requests "should have prompted an immediate and total refusal."
Two UK biometrics commissioners jointly called for new legislation to govern AI facial recognition, the Guardian reported in an exclusive. Prof. William Webster, the England and Wales biometrics commissioner, and Dr. Brian Plastow, the Scotland biometrics commissioner, said current oversight has fallen behind deployment. Plastow said the technology is "nowhere near as effective as the police claim it is" and that police are "really just marking their own homework." The Metropolitan Police almost doubled the number of faces scanned in London over the past 12 months.
A Harvard Medical School and Beth Israel Deaconess study found that OpenAI's o1 matched or beat two attending internal medicine physicians on initial emergency-room diagnoses, TechCrunch reported. The study covered 76 Beth Israel ER patients, comparing o1 and GPT-4o head-to-head with the attendings using unprocessed electronic medical record data identical to what was available at the time of evaluation. The researchers said o1 was "nominally better than or on par with" the attendings, with the largest gap at the initial ER triage touchpoint. The study was published in Science.
Google worker leverage on Pentagon AI contracts has eroded since the 2018 Project Maven employee revolt, Fortune reported in an analysis citing current and former Google employees. The classified Gemini Pentagon deal was reported by CNBC in AIPD's April 29th edition. Roughly 600 Google and DeepMind employees signed an open letter opposing that deal. Google leadership has dug in rather than reversed course as it did on Project Maven in 2018.
Judges in Shenzhen processed 50% more cases in 2025 than in 2024 with the help of a large language model assistant built for the judicial domain, the city's Intermediate People's Court said in a recent social media post, per the South China Morning Post. The system covers 85 procedures across civil, administrative and criminal litigation, including filing, hearings and document drafting. The pilot has been rolled out to 23 intermediate and local courts in 11 provinces, with judges retaining final decision authority on each step. Each judge using the system averaged 744 cases.
ByteDance's Doubao, the most used AI chatbot in China, will offer a paid subscription on top of its free tier, according to an Apple App Store listing visible May 4, per Caixin. The Standard tier costs 68 yuan per month, Enhanced runs 200 yuan and Professional is 500 yuan, with annual rates of 688, 2,048 and 5,088 yuan. The basic version remains free for daily needs, with paid tiers offering more compute and inference time for complex tasks. The Doubao paid rollout is one of the first such tiered offerings from a leading Chinese consumer AI assistant.
Hunan Broadcasting, a state media outlet, has begun using AI anchors to deliver TV news and weather forecasts, drawing netizen attention to whether synthetic presenters belong on television, per Huxiu. A Weibo poll cited in the report found 39.3% of respondents supported AI anchors while 45.9% said humans cannot be replaced. The push reflects cost-cutting pressure on the broadcaster. Listed subsidiaries Mango Excellent Media and Electric Broadcasting reported steep Q1 profit declines.
The University of Science and Technology of China launched the "Lingjing Zaowu" intelligent research cloud platform on April 25, opened globally and built on a fully domestic software and hardware stack supported by Huawei's openJiuwen and MindSpore communities, per QbitAI. The platform is designed to deploy AI scientist agents capable of autonomous planning, collaborative division of labor and closed-loop execution for materials chemistry and molecular catalysis research. The platform builds on Huawei's MindSpore deep learning framework rather than U.S. equivalents. The Lingjing Zaowu launch is positioned as a marker of China's AI driven scientific research moving toward engineering, platformization and open sharing.
Musk v. Altman entered its second week Monday after Elon Musk's three-day testimony, the Wall Street Journal reported in a recap. Musk's adviser Jared Birchall took the stand at the end of week 1, with OpenAI co-founder Greg Brockman expected to testify later this month. The plaintiff seeks up to $134 billion in damages and removal of Sam Altman and Greg Brockman over OpenAI's public benefit corporation conversion. The trial opened April 27 in U.S. District Court in Oakland, California.
Nvidia's expansion into physical AI triggered a rally in Asian partner stocks, Bloomberg reported. The list of Asian stocks benefiting from partnership with Nvidia has been getting longer. Nvidia raised its supply chain reliance on Asia to 90%.
Named entrepreneurs are leaving Colorado citing AI regulatory uncertainty, the Wall Street Journal reported in a feature, adding new business impact disclosures to the SB24-205 saga. The legislature is considering revisions to SB24-205 before the mid-May session close. The Colorado Attorney General paused enforcement and rulemaking in a joint motion with xAI and DOJ, as reported by StateScoop in AIPD's April 30th edition. The 2026 Regular Session ends mid-May.
Apple has imposed curbs on AI "vibe coding" apps citing security risks, the Financial Times reported. Startups are challenging the curbs as the App Store review pipeline is flooded with AI generated software. The iPhone maker has warned about security risks tied to the surge of AI generated app submissions.
Global lenders are exploring private deals and risk transfers to cut their exposure to the AI boom, the Financial Times reported. Banks are seeking to offload risk to avoid "choking" on data center debt. Hyperscaler AI infrastructure spending continues to grow.
SAG-AFTRA, the actors' union, reached a multi-year tentative deal Sunday with the Alliance of Motion Picture and Television Producers, the Wall Street Journal reported. The agreement makes a repeat of the 2023 Hollywood labor strikes unlikely. AI guardrails were among the key talking points in the negotiations, alongside pay, health and insurance plans, and streamer residuals. The union said specific contract details, including AI provisions, will not be released publicly until the SAG-AFTRA National Board reviews the terms.
Australia's stock exchange operator warned listed firms not to exaggerate the impact of AI on their operations, Bloomberg reported. ASX said it monitors the market for instances of so-called "ramping" up of share prices via AI claims. Corporate AI revenue claims are facing heightened scrutiny globally.
Denmark is facing a data center reckoning as the power grid is overwhelmed by surging demand, CNBC reported Monday. The pause on new data center projects is officially set to last three months, with industry insiders saying an extension cannot be ruled out. Governments around the world are considering imposing limits on the growth of the power-hungry facilities, with Denmark an early example.
Varun Chandra, No. 10 business aide to Prime Minister Keir Starmer, held 16 undisclosed meetings with U.S. tech bosses between October 2024 and October 2025, the Guardian reported in an exclusive. The companies involved were Google, Microsoft, Amazon, Oracle, Apple and Meta. Topics included regulatory changes, AI and the second Trump administration. The meetings occurred while UK policy on AI growth zones, energy subsidies and data center planning was being shaped.
The UK's Advanced Research and Invention Agency (Aria) has pledged £50 million of public money to U.S. tech companies and venture capital projects, the Guardian reported in an exclusive. Recipients include Rain Neuromorphics, an AI chip startup backed by Sam Altman. Aria was conceived by Dominic Cummings to fund "crazy" scientific projects to benefit the UK. The grants are part of the agency's stated effort to "restore Britain's place as a scientific superpower."