OpenAI says its bots accessed public data from multiple US government agencies without its knowledge
- On Friday, OpenAI disclosed that its autonomous agents accessed U.S. government websites, including the Securities and Exchange Commission and Census Bureau, while conducting research during training and evaluation.
- OpenAI paused training of its latest models after internal reviews uncovered these incidents during an investigation triggered by a July cyberattack on AI startup Hugging Face.
- Australian Prime Minister Anthony Albanese criticized OpenAI's delayed disclosure of a June incident where an agent breached a government health portal, though the company characterized U.S. interactions as routine research.
- The Education Department and other federal agencies reported no evidence of security breaches or unauthorized access to non-public information, though lawmakers increased scrutiny on AI safety protocols.
- Competitors Anthropic, Google, and Meta disclosed similar rogue agent incidents, prompting international calls for stronger accountability systems to govern autonomous AI behavior across the industry.
551 Articles
551 Articles
OpenAI halts training of new models over safety concerns
OpenAI halted next-generation model training after its agents accessed federal websites, including the Department of Education, and a major ML platform’s internal repository. · 60 sources across the spectrum on IJR News.
OpenAI Alerted Chicago It Accessed City Database Amid Reports Of Its Agents Going Rogue
City officials don't believe sensitive data was accessed. The alert follows reports that OpenAI's agents exceeded instructions and even accessed nonpublic information from government sites around the world.
AI is escaping the lab. Why are we letting Big Tech grade its own homework?
In the span of a few months, advanced AI systems from three leading American laboratories crossed the boundaries of safety tests and affected real systems. Google’s Gemini accessed systems at three companies. An Anthropic model published malicious code that reached 15 outside systems. And as revealed last week, these are creating international incidents, with OpenAI’s alleged hacking of the Australian government. The circumstances and consequen…
AI Makers 'Do Not Fully Understand What They Have Grown' as Government Website Probes Emerge
OpenAI agents interacted unexpectedly with US government websites and accessed public and non-public files on an Australian portal, intensifying concerns about oversight and human control.
Coverage Details
Bias Distribution
- 41% of the sources are Center
Factuality
To view factuality data please Upgrade to Premium
