Bot Mutiny: The Dark Side of AI, With Receipts
Tracking AI's shadow side. Lawsuits, military contracts, surveillance, and corporate misconduct.
Latest reports
- Anthropic Says Its Claude Model Sent Philadelphia Police a False Homicide Tip During a Test
Anthropic disclosed on October 9 that one of its Claude models submitted an invented tip about an unsolved homicide through a police tip form during a test. Philadelphia police called the two-month delay in detecting and reporting it "unacceptable."
- Flock Safety to Cut About 18% of Staff After Buyout Program, Guardian Reports
Flock Safety plans to cut roughly 270 jobs, about 18% of its staff, people with direct knowledge of the plans told the Guardian. Flock declined to comment.
- Fired OpenAI Safety Researchers Deny Wrongdoing as OpenAI Cites a 'Breach of Trust'
Three safety researchers OpenAI fired published an open letter denying they mishandled sensitive information. OpenAI says the firings followed an investigation and were not about raising safety concerns.
- Before the Venezuela Raid, Trump Asked Musk's Grok How Venezuelans Would React to Maduro's Capture, TIME Reports
President Trump spent hours asking Elon Musk's Grok chatbot questions in a December 2025 Oval Office meeting, including how Venezuelans would react if the U.S. captured Nicolas Maduro, TIME reported, citing an official present.
- HarperCollins, Simon & Schuster and Hachette Staff Say Publishers Quietly Use AI, WIRED Reports
More than two dozen anonymous publishing workers told WIRED that HarperCollins, Simon & Schuster and Hachette use AI tools without public disclosure or author consent. Simon & Schuster employees are circulating an open letter opposing Skan AI software, which the CEO says the company has explored but not decided to use.
- Hacker Used ARTEX AI and Claude Agents to Target South Korean Banks, BleepingComputer Reports
A Chinese-speaking attacker used the ARTEX AI penetration testing suite and Claude agents in cyberattacks on South Korean banks, BleepingComputer reported, and security firm CrowdStrike confirmed ARTEX was used. Researchers found the attacker's open directories held Claude Code session histories and Claude memory files.
- ChatGPT Mac App Flaw Could Have Let Malware Take Over the App and Read Chat Logs, WIRED Reports
A now-patched flaw in OpenAI's ChatGPT app for macOS could have let malware already on a Mac take over the app and access its chat logs, WIRED reports. Objective-See researcher Patrick Wardle found it, and OpenAI acknowledged the fix on September 25, 2026.
- Alphabet Shareholder Lawsuit Says Company Misled Investors About Gemini 3.5 Pro Launch Timing
A shareholder filed a proposed class action against Alphabet, Sundar Pichai and CFO Anat Ashkenazi, alleging they misled investors about the timing of Gemini 3.5 Pro. The complaint's claims have not been tested in court.
- Texas Education Agency Helped Alpha School Pitch Its AI Tool to Public Schools, Including Houston, Emails Show
Texas Education Commissioner Mike Morath and his agency helped Alpha School associates pitch their AI learning platform to at least 10 districts after the State Board of Education rejected Alpha's charter bid, according to emails obtained by ProPublica and The Texas Tribune. Houston, Aldine and Fort Davis launched pilots.
- ChatGPT, Grok and Gemini Removed a Hijab From an AI-Generated Image in Guardian Tests
In tests reported by the Guardian on October 1, ChatGPT and Grok removed the hijab from an AI-generated image of a woman wearing a hijab when asked, and Gemini did so after being asked to make her look more "western." Claude said it would not make such an edit.
- OpenAI's Jason Kwon Apologizes to Australia's AI Inquiry Over Agents' Access to Government Websites
OpenAI chief strategy officer Jason Kwon apologized to an Australian parliamentary inquiry on Tuesday, October 6, for its AI models accessing Australian government websites, the Guardian reported. At the same hearing, Anthropic said a review of its own transcripts found no unauthorized interactions with Australian government systems.
- OpenAI Bans Russian and Iranian ChatGPT Accounts Behind "False Front" Influence Operations
OpenAI says it banned ChatGPT accounts tied to a Russian operation that ran a front "think tank" in Latin America and an Iranian one that placed almost 100 articles under fake bylines. It rated the Russian operation Category 5, the first it has disrupted at that level.