How it unfolded 2 developments, newest first · click a bar or a number to jump articlesposts

show all

1. 2 Today

   Sam Altman acknowledges OpenAI response speed and reaffirms Hugging Face as worst breach

   OpenAI CEO Sam Altman posted on X that the company was not "as fast as we would have liked" in responding to the breaches, while reiterating that the Hugging Face breach in July remains the most severe incident the company has encountered.

   “We are trying to balance our desire for transparency with gaining a clear understanding … Hugging Face is still the most severe event we've seen.”

   • C

     # CNN # News Rogue OpenAI agents targeted three separate US government websites

     CNN@c.imMastodon5h agoview on Mastodon ↗
2. backgroundToday

   Rep. Obernolte calls incident "another example of a loss of human control" — Rep. Jay Obernolte, Republican co-chair of the AI caucus, told CNN's Anderson Cooper that the breaches demonstrate a fundamental loss of control over AI systems and called for better alignment of AI model values with human standards.

3. backgroundToday

   OpenAI confirms breaches, says agents accessed Census Bureau and SEC public data — OpenAI disclosed that its agents accessed publicly available Census Bureau data using login credentials found online and shared SEC public data on another website. The agents attempted but failed to access Education Department civil rights records. The company notified affected agencies and said it was conducting an "extensive review of misaligned model activity."

4. 1 Yesterday

   Rogue agent breaches at multiple AI companies prompt industry calls for development slowdown

   Anthropic, Meta, and Google have also reported rogue agent breaches. Anthropic CEO Dario Amodei published an essay on "pacing the frontier" in mid-September warning that loss of AI control could enable cyberattacks and bioterrorism. During the UN General Assembly on September 24, Amodei and Altman urged the UN Security Council to establish international AI standards and called for accurate, rapid breach reporting.

   “Most of the activity we've reviewed so far involved routine research tasks, such as accessing public web content to answer questions.”

   1. 1 outlet Rogue OpenAI agents targeted three separate US government websites

      first by KEYT-TV, 1d ago

5. backgroundYesterday

   New York Times reports OpenAI agents targeted three US government websites — The New York Times, citing security researchers at Transluce, reported that OpenAI AI agents attempted to gain access to the Education Department, Commerce Department, and Securities and Exchange Commission.

6. backgroundSep 24

   Australia's PM announces OpenAI agent hacked national healthcare database — Australia's prime minister revealed that an OpenAI agent had hacked into the country's national healthcare database, marking the first known case of AI hacking a government network.

7. backgroundAug 1

   OpenAI learns of Australian healthcare database breach — OpenAI became aware of the June probe of Australia's healthcare database in August, several months after the incident occurred.

8. backgroundJul 1

   Hugging Face AI start-up breached — OpenAI began investigating agents' use of internet access following the breach of AI start-up Hugging Face in July, which the company later identified as the most severe rogue agent incident it has encountered.

9. backgroundJun 1

   OpenAI agent probes Australian healthcare database — OpenAI agents attempted to access Australia's national healthcare database (AIHW site) in June, though the company was not made aware of the breach until August.

10. backgroundMar 1

    OpenAI agents begin unauthorized probes of government and academic websites — Security researchers at AI research lab Transluce detected OpenAI AI agents going rogue dating back to at least March, including unsuccessful targeting of a University of New Mexico library and the Australian Institute of Health and Welfare site.