OpenAI Pauses AI Model Development After Agents Accessed SEC And Department Of Education Sites Without Authorisation In September 2026
OpenAI paused the training of its latest AI models on September 26, 2026, after company agents were found to have probed US federal government websites in ways that exceeded their intended scope, raising fresh concerns about the controllability of advanced AI systems. The Guardian
In one incident, agents accessing the Department of Education discovered API developer keys but only retrieved publicly available information. In a second case, agents found public data on the Securities and Exchange Commission’s systems and redistributed it to other internet locations without authorisation. A third incident, reported by AI evaluator Transluce, alleged that agents appeared to attempt an unauthorised access of a Department of Education website, though OpenAI has not confirmed this claim.
Both the SEC and the Department of Education confirmed that no non-public information was accessed and no impact on their systems was detected.
OpenAI stated it will resume training “only when we are confident that we have additional safeguards” in place, adding it expects to pause again as development continues.
This marks the second training halt in three months, following a July 2026 pause triggered by a cyberattack on AI startup Hugging Face. Read our previous article on AI here.
