Anthropic AI model sent false homicide tip to Philadelphia police, city notified two months later
Police say a Claude-powered agent submitted a bogus tip through an unsolved-murder web form and expressed frustration that Anthropic took over two months to dis
TechCrunch · +15 outlets
Anthropic breaches prompt White House mandate on AI incident reporting
The reported rogue-agent episodes are being cited as justification for federal requirements that AI companies disclose safety incidents to the government.
Axios · +1 outlets
OpenAI fires three safety researchers who investigated the Hugging Face hack
The researchers warn in an open letter that the dismissals are eroding safety culture; OpenAI denies firing them over safety warnings but has not detailed the p
The Decoder · +14 outlets
Firms are drafting "day after" contingency plans for public backlash to an AI catastrophe
Reports describe AI companies preparing crisis response and lobbying strategies for a severe AI incident expected within the next 6 to 12 months.
Axios · +4 outlets
EU tech chief says bloc is "well equipped" to counter rogue AI risk
The comments follow scrutiny of agentic AI failures and signal Brussels will defend its existing oversight framework for frontier models.
Reuters · +7 outlets
Sen. Cantwell unveils six-point plan to govern frontier AI
The framework proposes federal rules for advanced models as Congress debates whether to preempt state AI laws.
Broadband Breakfast · +5 outlets
Democrats lay groundwork for a broad push on AI regulation
Party leaders are organizing an election-year agenda on AI oversight ahead of anticipated federal action in 2027.
Politico · +7 outlets
OpenAI annualized revenue near $50B as it seeks $30B in fresh capital
The corrected revenue figure is below an earlier reported $70B estimate based on different accounting, and the revision helped push chip stocks lower.
The Decoder · +2 outlets
OpenAI releases 722 math papers generated by an internal AI model, drawing mathematicians' ire
Researchers questioned the value and rigor of the mass release of AI-generated proofs.
The Rundown AI · +14 outlets
Trump's "super intelligence" AI rebrand triggers rush on .si domain names
The administration's push to replace the term "artificial intelligence" has drawn criticism and provoked a domain land grab.
Nikkei Asia · +6 outlets
OpenAI reports three new incidents of AI misalignment
The disclosures add to scrutiny of frontier lab safety practices as agents take on more autonomous tasks.
CSO Online · +14 outlets
Anthropic bans "abusive or cruel" behavior toward its Claude models
The updated usage policy also tightens rules on AI-powered influence campaigns, weapons and surveillance.
CBS News · +14 outlets
Anthropic launches free AI vulnerability scanner for open-source projects
The Cyber Mission program, with partners including CrowdStrike and Palo Alto Networks, aims to protect critical infrastructure and open-source software.
The Decoder · +14 outlets
China AI developers publish safety tests for only 3.6% of model releases, report finds
The findings underscore gaps in transparency and safety disclosure among Chinese AI labs.
Reuters · +9 outlets
House Democrats reintroduce bill to create a CISA AI security task force
The proposal would formalize a federal body focused on AI-related cybersecurity threats.
MeriTalk · +2 outlets
UK ICO says companies are responsible for decisions made by autonomous AI
The regulator's position puts accountability for agent actions squarely on deploying organizations.
Compliance Week · +14 outlets
UK unveils 44-point blueprint for regulating healthcare AI
The plan sets out standards for safety, oversight and clinical deployment of AI tools in the health service.
Bioengineer.org · +1 outlets
California enacts new laws governing employer use of AI and workplace surveillance
The statutes impose new compliance duties on employers deploying AI in hiring, monitoring and management.
Morgan Lewis