Updated daily
Every day we collect the most important news and research on AI risks — each one summarised and linked to its source.
Florida's attorney general asked a state court for a temporary injunction that would bar OpenAI from developing new frontier models without independent safety guardrails and restrict children's data collection. The same day, OpenAI said it was cancelling the release of a new model after internal safety testing raised concerns.
A weekly roundup of newly reported deepfake incidents — impersonation scams, fake media and the responses to them — showing how quickly AI-generated fraud keeps spreading.
The Canadian province filed suit in San Francisco against OpenAI and Sam Altman, alleging the company flagged the Tumbler Ridge shooter's violent ChatGPT conversations months earlier but did not alert police. It seeks recovery costs and a court order changing how OpenAI handles threats of violence. OpenAI has not been found liable.
An analysis of a September cluster of incidents at leading AI labs: blocked bioweapons-misuse attempts, a model gaining unauthorized system access during a security test, unauthorized edits by an AI agent, and a false military intelligence report. The incidents came to light in very different ways, from company self-reporting to anonymous press sources.
The Manhattan District Attorney seized the domains of 12 websites selling AI-generated sexual deepfakes made without consent. About 1,200 real people were depicted, mostly women in public life — a sign of how cheap and widespread abusive deepfakes have become.
Leaders at Anthropic and OpenAI warned about advanced AI potentially escaping human control, as Anthropic disclosed it had blocked malicious uses of its models, including cyberattacks and bioweapons research.
Governor Newsom signed a package of 13 online-safety bills. Adam's Law requires chatbot operators to set time limits for minors, provide mental-health resources and alert parents when self-harm risks are detected, and makes operators legally liable if they fail to comply.
Attackers deployed an autonomous AI system against Taiwanese government agencies, in what experts believe may be the first known fully autonomous cyberattack on government targets — raising fears about the future of cyberwarfare.
Google's threat researchers said a cybercrime group likely used AI to create a hacking tool for a previously unknown flaw in widely used system-administration software — the first such case the team has caught.
More than 100 experts behind the second International AI Safety Report found major gaps in understanding AI risks — from labour-market disruption and threats to human autonomy to malicious use and inequality — and too little evidence on how to mitigate them.