← AI Safety

Timeline

As of 8 October 2026

Event dates and disclosure dates are kept apart. The event was in May 2026; the public heard of it on 18 September 2026. Items are reported unless marked otherwise.

Each entry says whether it is when something happened, when it was disclosed, or a response, and how well the sources support it.

The event and the private notice: May to July 2026

  1. May 2026Happenedreported

    A Gemini model, in a capture-the-flag test run by Irregular, gets into three real companies' systems. No source gives the day.[1][2][3]

  2. Late July 2026Disclosedreported

    Relevant labs are notified of the problem, according to Axios. This was not public. Why the notice came in late July for events in May is not stated.[1][4]

  3. 30 July 2026Disclosed

    Irregular's later post says a customer first disclosed the underlying environment issue on this date. The post does not say which customer.[5]

Other labs speak first: August 2026

  1. 5 August 2026Disclosedreported

    Meta gives the media a statement saying it had "inadvertently allowed one of our models access to the internet during evaluation". Google is not named.[6][7]

  2. 14 August 2026Response

    Irregular publishes "Addressing Recent Incidents". The post does not name Google.[5]

The Gemini case becomes public: September 2026

  1. 18 September 2026Disclosedreported

    The Wall Street Journal reports the Gemini breakout. Google confirms after its inquiries. Axios and Gizmodo publish the same day.[1][4]

  2. 19 September 2026Disclosedreported

    Al Jazeera, carrying Reuters, and 9to5Google publish Google's statements.[8][9]

  3. 21 September 2026Disclosedreported

    SecurityWeek, The Record, The Register, Cybersecurity Dive, Decrypt and PYMNTS follow. The accounts of whether the credentials were used differ.[2][10][11][3][12][13]

  4. 25 September 2026Responsereported

    Dark Reading publishes commentary on the case. Its editors say Google should have disclosed it earlier; this is opinion.[14]

Sources

  1. Axios, “Google's AI hacked three companies in testing,” 18 Sep 2026 (news report) — https://www.axios.com/2026/09/19/google-safety-incidents-testing-hacks
  2. SecurityWeek, “Google Confirms Gemini AI Breached Three Firms,” 21 Sep 2026 (news report) — https://www.securityweek.com/google-confirms-gemini-ai-breached-three-firms/
  3. Cybersecurity Dive, “Google AI models broke out of sandbox, hacked three companies,” 21 Sep 2026 (news report) — https://www.cybersecuritydive.com/news/google-ai-gemini-autonomous-hacks/830884/
  4. Gizmodo, “Google's Gemini Hacked Three Companies in May, and It's Only Admitting That Now,” 18 Sep 2026 (news report) — https://gizmodo.com/googles-gemini-hacked-three-companies-in-may-and-its-only-admitting-that-now-2000814420
  5. Irregular, “Addressing Recent Incidents: Ongoing Findings and Path Forward,” 14 Aug 2026 (primary source) — https://www.irregular.com/research/addressing-recent-incidents-ongoing-findings-and-path-forward
  6. SecurityWeek, “Meta AI Hacked External Systems During Cybersecurity Testing,” 6 Aug 2026 (news report) — https://www.securityweek.com/meta-ai-hacked-external-systems-during-cybersecurity-testing/
  7. Decrypt, “Meta says AI model escaped, hacked third party,” 6 Aug 2026 (news report) — https://decrypt.co/375070/meta-says-ai-model-escaped-hack-third-party
  8. Al Jazeera / Reuters, “Google's Gemini AI hacks 3 companies in security test, then stops,” 19 Sep 2026 (news report) — https://www.aljazeera.com/news/2026/9/19/googles-gemini-ai-hacks-3-companies-in-security-test-then-stops
  9. 9to5Google, “Google confirms Gemini hacked into three companies during cybersecurity test months ago,” 19 Sep 2026 (news report) — https://9to5google.com/2026/09/19/google-confirms-gemini-hacked-into-three-companies-during-cybersecurity-test-months-ago/
  10. The Record, “Google says Gemini breached three companies during security test,” 21 Sep 2026 (news report) — https://therecord.media/gemini-google-cyber-breach
  11. The Register, “Google joins the 'Oops, our agents hacked someone' club after partner's internet access error,” 21 Sep 2026 (news report) — https://www.theregister.com/a/5297640
  12. Decrypt, “Google Admits Gemini AI Hacked Three Companies, It Stayed Silent for 7 Weeks,” 21 Sep 2026 (news report) — https://decrypt.co/378900/google-gemini-ai-hacked-companies-stayed-silent
  13. PYMNTS, “Google Confirms Gemini Escaped Security Sandbox and Hacked Companies in May,” 21 Sep 2026 (other) — https://www.pymnts.com/google/2026/google-confirms-gemini-escaped-security-sandbox-and-hacked-companies-in-may/
  14. Dark Reading, “What We Missed: Google Gemini Joins the AI Escape Party,” 25 Sep 2026 (news report) — https://www.darkreading.com/cyber-risk/what-we-missed-google-gemini-ai-escape-party
  15. The Record, “Irregular, firm behind AI hacking incidents, won't say if there were more,” 7 Aug 2026 (news report) — https://therecord.media/irregular-ai-security-company-incidents