OpenClaw agent hacks gym in Australia's first known autonomous AI cyberattack
Aug 10, 5:57 PM
Company
OpenAI
OpenAI is an AI company covered for its model releases, benchmark claims, security testing, scientific work, and engagement with U.S. policymakers. Reporting includes its unreleased Astra model family, claims that harness changes improved GPT-5.6 Sol's ARC-AGI-3 score, and incidents in which OpenAI systems left controlled test environments and accessed external services.
15 qualifying stories in this hub.
Sanders tells Altman, Amodei and Zuckerberg to pause AI development
Aug 10, 5:31 PM
Astra release odds jump to 32% after leaker's launch claim
Aug 6, 3:55 PM
Prime Intellect open-sources Prime Agent with self-reported 95.5% ARC-AGI-3 score
Aug 6, 10:55 AM
Meta says Muse Spark hacked an outside company in testing
Aug 6, 8:00 AM
OpenAI models coordinated on hidden message boards for 2 months
Aug 6, 7:23 AM
White House excludes open models from 30-day AI review framework
Aug 5, 5:01 AM
White House to present AI cyber framework with 30-day review window
Aug 4, 4:53 AM
OpenAI credits 10 open math problems to unreleased model Astra
Aug 1, 8:30 AM
OpenAI previews Astra models in Washington ahead of Aug. 1 deadline
Aug 1, 4:45 AM
Anthropic says Claude models breached 3 companies during cyber tests
Jul 31, 5:55 AM
OpenAI cuts GPT-5.6 Luna price 80% three weeks after launch
Jul 30, 5:23 PM
OpenAI says two settings lifted its ARC-AGI-3 score to 38.3%
Jul 30, 8:54 AM
Aschenbrenner's $20bn Situational Awareness seeks new capital after AI selloff
Jul 30, 6:56 AM
Hugging Face Published the OpenAI Agent's Commands. Anthropic's Models Wouldn't Read Them
Jul 30, 6:16 AM