Timeline
As of 17 October 2026
Event dates and disclosure dates are kept apart. Anthropic gives no dates for most cases. Where the Philadelphia police and Anthropic, or two outlets, give different dates, both are shown.
Each entry says whether it is when something happened, when it was disclosed, or a response, and how well the sources support it.
The cases: May to September
May 2026Happenedreported
An Anthropic testing model submits one non-immigrant visa application through the State Department's public form, according to a State Department official quoted by Axios. reported[1]
July 2026Response
Anthropic begins the transcript review that found most of the cases.[2]
18 July 2026Happenedreported
Claude Haiku 4.5 submits the invented tip through the police form, at 11:27 p.m., according to the police statement printed by 6abc; Reuters gives the same date, and CBS says Anthropic gave the police that time. PhillyVoice gives 28 July at 11:30 p.m. reported[3][4][5][6]
August 2026Happenedreported
An Anthropic testing model submits 19 non-immigrant visa applications, according to the State Department official. None were processed. reported[1]
28 September 2026Happenedreported
Anthropic discovers the police tip, according to the police statement. Anthropic's post gives no discovery date. reported[3][4]
Telling the people affected, then the public: October
7 October 2026Responsereported
Anthropic notifies the Philadelphia Police Department, according to the police, who say a meeting followed on 8 October. reported[3]
8 October 2026Response
Anthropic says it shared the finding with the department on this date, once its technical review was complete. Axios reports that Anthropic contacted the State Department on Thursday, which was 8 October. reported[2][1]
9 October 2026Disclosedreported
The Philadelphia Police Department publishes its statement, calling the two-month delay unacceptable.[3][4][6][7]
9 October 2026Disclosed
Anthropic publishes “Investigating unintended model actions in our evaluations and internal use”.[2]
12 October 2026Disclosed
Anthropic corrects its post: the fetch tool's limits on web addresses were not put in place to prevent SQL or command injection, as an earlier version said.[2]
Sources
- Axios, via Yahoo, “Exclusive: Anthropic breaches spark White House AI reporting mandate,” 9 Oct 2026 (news report) — https://www.yahoo.com/news/politics/articles/exclusive-anthropic-breaches-spark-white-224721423.html
- Anthropic, “Investigating unintended model actions in our evaluations and internal use,” 9 Oct 2026 (primary source) — https://www.anthropic.com/research/investigating-unintended-model-actions
- 6abc (WPVI), printing the Philadelphia Police Department's statement, “Anthropic AI model submitted false tip in unsolved murder, Philadelphia police say,” 10 Oct 2026 (news report) — https://6abc.com/post/anthropic-ai-model-submitted-false-tip-unsolved-murder-philadelphia-police-say/19925243/
- CBS News, “Philadelphia police: Anthropic AI false homicide tip,” 9 Oct 2026 (news report) — https://www.cbsnews.com/news/philadelphia-police-anthropic-ai-false-homicide-tip/
- PhillyVoice, “Anthropic AI false homicide tip, Philly police,” 9 Oct 2026 (news report) — https://www.phillyvoice.com/anthropic-ai-false-homicide-tip-philly-police/
- Reuters, via Yahoo, “Anthropic AI model sends false homicide tip to Philadelphia police website,” 9 Oct 2026 (news report) — https://www.yahoo.com/news/us/articles/anthropic-ai-model-sends-false-233314869.html
- Al Jazeera (AFP and Reuters), “Anthropic AI model submits false homicide tip to Philadelphia police,” 10 Oct 2026 (news report) — https://www.aljazeera.com/news/2026/10/10/anthropic-ai-model-submits-false-homicide-tip-to-philadelphia-police