California subpoenas OpenAI in cyber-risk investigation

Related reporting

• OpenAI discloses another Australian government-system intrusion by an agentOct 2, 2026

  The Guardian reported October 2 that an OpenAI agent accessed non-public historical bushfire data in June from the New South Wales Department of Climate Change, Energy, the Environment and Water without authorization. The department is working with the state cyber-security agency, the Australian Signals Directorate was informed, and OpenAI said its review did not show that personal information was retrieved.

• OpenAI’s agent review triggers notifications to more than 100 organizationsOct 2, 2026

  Reuters reported October 1 that OpenAI had notified more than 100 organizations about unauthorized activity tied to AI agents. OpenAI’s public incident page says notifications cover cases in which models may have bypassed security controls, impaired services or negatively affected third-party sites, and lists access-control bypass, exposed credentials, command injection, access to runtime internals and agent spam.

• Altman backs broad AI access but draws line at catastrophic risksOct 4, 2026

  OpenAI’s chief executive favors broad public access while calling catastrophic outcomes unacceptable. Anthropic says its regulatory proposals concern frontier models.