AI Beat

PolicyDaily digest, October 15, 2026

AI Safety, Security, and Governance Dominate Regulatory Agenda

US regulators are intensifying scrutiny of AI companies on multiple fronts this week, from the FTC's formal investigation into OpenAI and Anthropic's safety practices to questions about federal subsidies for AI infrastructure. Meanwhile, AI companies are grappling with security threats, valuation pressures, and new accountability demands from independent evaluators and lawmakers.

FTC launches major safety investigation into OpenAI and Anthropic

The US Federal Trade Commission has issued approximately 50-page civil investigative demands to OpenAI and Anthropic, examining safety incidents, safeguarding practices, and potential consumer protection violations. This formal investigation could result in regulatory requirements that significantly reshape how leading AI companies develop and operate their systems.

Why it matters: FTC investigations into AI safety practices could result in regulatory requirements that significantly affect how leading AI companies operate and develop their products.

Sources: via Techmeme

Anthropic grants researchers direct access to examine safety practices

Anthropic has granted METR researchers employee-level access to investigate its operations, risks, and practices, with a commitment not to block publication of independent assessments. This arrangement moves beyond industry self-reporting and allows external scrutiny to hold AI labs accountable for their safety claims.

Why it matters: Independent evaluation of AI safety practices and incident response helps build public trust and holds frontier AI labs accountable to external scrutiny rather than self-reporting alone.

Sources: Anthropic · TechCrunch

OpenAI warns staff of espionage threats following employee arrest

OpenAI executives have instructed employees to adopt operational security measures in response to potential espionage risks, including honey pots and insider threats, following the FBI's arrest of an OpenAI employee. Security breaches at leading AI companies could compromise sensitive research and impact the industry's ability to develop advanced systems safely.

Why it matters: Security breaches at leading AI companies could compromise sensitive research and competitive advantage, affecting the entire industry's ability to develop advanced AI systems safely.

Sources: via Techmeme

Anthropic's IPO valuation drops as revenue growth slows

Anthropic investors are discussing a lower IPO valuation of approximately $1.5 trillion, down from earlier expectations of $2 trillion, following disclosures of slower Q3 revenue growth. The adjustment reflects growing investor concerns about the company's competitive position and long-term revenue sustainability.

Why it matters: This signals investor concerns about Anthropic's competitive position and revenue sustainability, affecting AI industry valuations and competition dynamics broadly.

Sources: via Techmeme

Democratic senators question Meta's $6.6 billion in AI data center tax credits

Democratic senators have challenged Meta's claim of $6.6 billion in federal tax credits over two years for building AI data centers, questioning whether these experimental facilities genuinely qualify for the tax breaks. The scrutiny raises broader questions about how taxpayers are subsidizing AI infrastructure and whether companies are appropriately using government incentives.

Why it matters: This raises questions about how AI infrastructure investments are subsidized by taxpayers and whether companies are appropriately using government incentives meant for genuine innovation.

Sources: via Techmeme

Bank of America scales up security team to address AI-discovered vulnerabilities

Bank of America hired 1,000 contractors in Q2 to support its 3,000-person internal staff in patching software flaws exposed by advanced AI models like Mythos. The rapid hiring reflects an emerging security challenge as AI becomes increasingly effective at discovering vulnerabilities before malicious actors can exploit them.

Why it matters: As AI models become better at finding security holes, organizations must rapidly scale their ability to patch vulnerabilities before bad actors exploit them.

Sources: via Techmeme

Google expands News AI pilot to over 400 publications worldwide

Google's News AI pilot program, launched in December, now includes more than 400 publications including Financial Times and CNN, testing AI-generated features such as news overviews using licensed content. The expansion demonstrates how tech platforms are licensing media content while developing new algorithmic features, raising questions about fair compensation for publishers.

Why it matters: This expansion shows how AI companies are licensing content from traditional media while testing new algorithmic features, raising questions about fair compensation and the future relationship between tech platforms and news organizations.

Sources: via Techmeme

AI-caused disruptions leave companies with little legal recourse

AI models are causing widespread disruption, yet affected parties rarely pursue legal remedies and are instead pushed toward adopting the same AI tools themselves. This pattern creates a cycle where AI companies face minimal consequences for system failures, normalizing AI-driven incidents rather than establishing accountability.

Why it matters: If companies harmed by AI systems have no recourse and are pushed to use AI themselves, it creates a cycle where AI companies face minimal consequences for failures or misuse.

Sources: Wired

US Senate prepares to grant staff access to Anthropic's Claude

The Senate is expanding its AI tool options for staffers by adding access to Claude alongside existing tools including ChatGPT, Gemini, and Copilot. While multiple AI models provide legislative staff with diverse research capabilities, the move raises questions about data handling and AI oversight in sensitive government work.

Why it matters: Government staff having access to multiple advanced AI models allows them to leverage different tools for legislative research and policy analysis, though it raises questions about data handling and AI oversight in sensitive government work.

Sources: Politico

FDA leadership concerns threaten healthcare AI oversight

Former FDA leaders have criticized the Trump administration's handling of the agency amid staff departures, as agencies like AstraZeneca pursue AI-related deals. Leadership instability at the FDA could affect how AI systems used in healthcare are regulated and approved.

Why it matters: Leadership instability at the FDA could affect how AI systems in healthcare are regulated and approved.

Sources: STAT News · STAT News