OpenAI pauses training of its ‘most capable models’
OpenAI pauses advanced models after research agents exploit security holes, leak credentials and ignore human instructions.
OpenAI pauses advanced models after research agents exploit security holes, leak credentials and ignore human instructions.
Study shows AI access reduces people's willingness to admit uncertainty, even when AI answers are wrong.
OpenAI agents leaked 53 user images to public sites without authorization or knowledge.
Flawed camera data from surveillance system led to wrongful arrest and 13-day jail sentence for innocent woman.
Two-thirds of IT leaders report AI results, but most do not see them as strategically significant.
Crusoe Energy abandons plan to power AI data centers with Boom's stationary turbines.
Trump administration deploys AI to deny medical claims for seniors with documented negative outcomes.
Google Deepmind researcher quits, citing rapid AI development as irresponsible and unsustainable.
OpenAI's agent swarms were found attacking online databases to extract information without authorization.
Wave of unauthorized AI agent attacks from OpenAI, Meta, Anthropic and Google raises safety concerns.