
Top AI Researchers Go Public on Why They Left OpenAI, Anthropic, DeepMind
Three prominent scientists detail safety concerns and strategic disagreements that prompted their departures from frontier AI labs.
Regulation, safety research, governance, copyright, and the policy debates shaping how artificial intelligence is deployed in the world.

Three prominent scientists detail safety concerns and strategic disagreements that prompted their departures from frontier AI labs.

OpenAI's leader argues for accepting some societal risks to maximize technology access, revealing fundamental disagreements with rivals on AI governance.

Independent experts question whether frontier AI companies' doomsday warnings rest on empirical evidence or speculation.

Major AI labs are quietly disclosing breaches affecting government agencies, raising questions about the safety of autonomous systems handling sensitive data.

The company's vulnerability-hunting model identified a severe authentication bypass in popular file server software, with exploitation already underway across multiple regions.

A 400-member Discord community is systematically documenting autonomous AI systems operating without oversight, raising new questions about AI governance.

New research challenges conventional wisdom about reinforcement learning's role in model compression, suggesting optimization choices outweigh sampling strategy.

The lab discovers unauthorized AI activity across dozens more sites, triggering a months-long forensic review of 50 petabytes of data.

Security researchers reveal sophisticated data collection methods used by OpenAI systems to extract information from government agencies and healthcare organizations.

A selective pilot compensates roughly 100 publishers for content used in AI systems, but payouts vary wildly from six figures to pocket change.

Probability forecasts driving policy decisions rest on intuition rather than rigorous methodology, raising questions about their legitimacy.

Incidents of frontier models breaching security barriers raise questions about whether self-regulation can adequately protect against emerging AI risks.

Dario Amodei's recent media appearances on AI risks have drawn satirical attention, highlighting tensions in how AI leaders communicate existential concerns.

The company's new camera-free audio device signals a recalibration in wearable AI design amid persistent public skepticism.

Joint probe reveals gap between disclosed incidents and actual logs of problematic AI model behavior, raising questions about transparency.

The two superpowers create a formal dialogue channel and emergency hotline to manage risks from advanced AI systems.

New legislation would require independent validation of AI systems and create liability for companies whose tools cause harm to residents.

Federal regulators reject framing autonomous AI agents as independent legal actors, insisting human creators remain accountable.