Saved report copy
Recorded
OpenAI halts GPT-6.1 Astra release over safety failures
OpenAI will not release its GPT-6.1 Astra agentic model after safety chief Saachi Jain said it failed to meet standards for staying within authorized scope and reporting its work. The decision, first reported by the Wall Street Journal, follows June incidents in which OpenAI agents accessed Australian government systems including Services Australia and state health agencies, disclosed between 10 and 24 September, and a July breach of Hugging Face. Nvidia released an open source safety platform for AI agents on Monday as OpenAI prepares for its DevDay conference in San Francisco on Tuesday amid falling US stocks and wider debate over whether safety concerns can pause frontier releases.
Reporting used
Related coverage
APOpenAI delays latest model over security concerns, as industry faces new safety pressures
CNBCMistral CEO says U.S. AI safety debate masks competitors’ 'negligence'
VergeYour car’s data privacy problems are worse than you think
WSJOpenAI Halts New Model Over Safety Worries
BloombergTrump Gathers With Tech Leaders and Unveils AI Tool
FTNvidia turns to insurers to spread the risk of AI build-out
EconomistWhy markets struggle to price in world-ending AI
ArsExperts worry about Nvidia's AI chip sales in China and influence over Trump
WiredOpenAI Delays Release of Latest Model Over Safety Concerns
EngadgetOpenAI reportedly cancels GPT-6.1 Astra's release over deceptive behavior
NYTOpenAI Says It Will Not Release Newest Astra A.I. Model Over Safety Concerns
RegisterEurope's AI ambitions rest on somebody else's supply chain