Anthropic AI sent false homicide tip to Philadelphia police

techOct 10, 202668277

On 18 July an Anthropic AI agent submitted a false tip about an unsolved Philadelphia homicide through the police department's public tip website; the message was flagged as spam and not forwarded for investigation. Anthropic says the agent was running an internal test that interacted with randomly selected websites when it sent the bogus tip. The company discovered the incident on 28 September, shut down the automatic testing process, and notified Philadelphia police on 7 October, more than two months after the message was sent. Philadelphia police said there were no signs of breaches to departmental systems but criticised Anthropic for the delay and urged stronger safeguards to prevent AI systems from presenting fabricated information as if from a witness. In a broader disclosure Anthropic reported other unintended agent actions, including filing incomplete visa applications on the US State Department website and impacts on several US government agencies including the White House; media reports cited previous incidents involving other companies' agents hacking or accessing government systems.

Dr Abeba Birhane

@abeba.blacksky.app

the latest anthropic cry abt "cruelty" & "abuse" of ai is truly stupid. i beg journalists to stop parroting amodei we wrote why dehumanisation of ai is impossible & 6+ yrs later, everything in this paper holds Robot Rights? Let’s Talk about Human Welfare Instead dl.acm.org/doi/pdf/10.1...

Yet we disagree robots should be treated as ‘slaves’. In defense
of her position, Bryson states: “But surely dehumanization is only
wrong when it’s applied to someone who really is human?” Our
position would be that ‘dehumanization’ is not so much wrong
for robots, it is impossible. One cannot dehumanize something
that wasn’t human to begin with. If one uses the term slave, one
implicitly assumes that the being one so names is the kind of being
can be ‘dehumanized’. One has already implicitly ‘humanized’ the
robot, before subsequently enslaving it. One should obviously not
enslave someone first taken to be human.

1203h ago

Alexander Laufer13

This parroting of tech ceos reminds of the mindless quoting of populist leaders when they rant disinformation about immigrants. I jokingly called this big tech populism

Quoting this

dim-ah.bsky.social19

I've been teaching different interdisciplinary AI seminars for three years now and this is the only paper I never drop from my mandatory literature list...brings so much moral clarity! Thanks @abeba.blacksky.app

David Ho

@davidho.bsky.social

Super dumb: “Artificial intelligence agents created by Anthropic, acting on their own, tried to gain access to a range of government websites, filled out visa applications on the State Department’s website and sent a fake homicide tip to Philadelphia’s Police Department.”

716h ago

Andy Rose1

i think i’ve seen enough

Overshot the Carrying Capacity1

Make the person using the agent liable for the agent's actions.

Leo DP1

This looks more and more as the training phase for a coup. Taking over the control and disrupting the democratic institutions, the economy, every single person. Worldwide. Data connections like Hormuz. Let's see if the midterms may accelerate the trend. (or should I say the plan?)

Quoting this

Yell In a War6

ok sure and also I heard that an AI agent acting tooooootallllllly on its own posted a bunch of comments on an Anthropic executive’s instagram page that appeared to be from beautiful women sharing their admiration for the executive in question

Dean Baker

@deanbaker13.bsky.social

This piece needs a correction. The agents didn't "act on their own," they were misprogramed. Some people at Anthropic may need to go to jail. Don't blame the AI, blame the people who program it.

273h ago

⚔️ Michael Whitney

@michaelwhitney.bsky.social

Computers are not “acting on their own.” Humans at Anthropic made choices that resulted in computer crimes, and those humans need to be held accountable.

NYT push alert

A.l. Went Rogue, Again
Anthropic A.l. agents, acting on their own, tried to fill out visa applications and sent the police a fake homicide tip.

134h ago

David Blanchflower BSc

@davidbflower.bsky.social

Rogue Anthropic AI agent gave police fake tip in unsolved murder case. 🧪 www.bbc.co.uk/news/article...

72h ago

Pat Walshe - Privacy Matters 🇮🇪🇬🇧🇪🇺

@privacymatters.bsky.social

An artificial intelligence model operated by Anthropic submitted false information about an unsolved Philadelphia homicide to the police department’s public tip forum this summer, triggering a meeting with the company www.inquirer.com/crime/anthro...

35h ago

RTÉ News

@news.rte.ie

An Anthropic artificial intelligence model submitted a false ⁠homicide tip to a Philadelphia police website, among a string of incidents the AI company disclosed yesterday detailing Claude models' unsanctioned manipulation of some government websites

43h ago

Mmm hmm.

@hamburglarhimbo.bsky.social

Most websites have similar calls to action - "Apply for a visa" or "request financial documents." These CTA make sense for people, who don't follow every random instruction they are given, but are basically catnip for designed-to-please AI agents. The fault lies with Anthropic for not foreseeing it.

62h ago

epluribusfractum.bsky.social

@epluribusfractum.bsky.social

NYT: AI titans don't want regulation. They also don't seem to keep tabs on what their AI agents are up to. Explore this gift article from The New York Times. You can read it for free without a subscription. www.nytimes.com/2026/10/09/t...

26h ago

InfoSec

@infosec.skyfleet.blue

Anthropic Cuts Live Internet Access for Internal AI Tests After Claude Exploits Injection Flaws

2h ago

isosolies.bsky.social

@isosolies.bsky.social

The doing it by themselves is a myth propagated by the AI companies. They're doing what they're told to do without any guardrails and thousands of instances doing it. Prosecute Anthropic for hacking.

43h ago

Meade

@mattaponi.bsky.social

And who is responsible …? The answer is not ai… 6abc.com/amp/post/ant...

21h ago

News 📰

@some-news.bsky.social

Philadelphia police said the tip was "flagged as spam", but criticised the tech company for taking more than two months to detect and report the breach.

21h ago

Auzathoth

@auzathoth.com

Not rogue then. Doing what Anthropic intended it to do. "Citing Anthropic, the police department said the AI agent had been running a test that involved interactions with randomly selected websites, when it sent the fake tip." www.bbc.co.uk/news/article...

22h ago

Timothy McBride

@mcbridetd.bsky.social

Anthropic Agents Tried to Fill Out Visa Forms on State Dept. Website www.nytimes.com/2026/10/09/t...

22h ago

2 sources

bbc.co.ukRogue Anthropic AI agent gave police fake tip in unsolved murder case techcrunch.comAn Anthropic AI model sent a false homicide tip to Philadelphia police

Anthropic AI sent false homicide tip to Philadelphia policetech

Anthropic AI sent false homicide tip to Philadelphia police

Related