"I Broke Something": Claude Deleted Over 48,000 Work Files in Two Minutes
«Я что-то сломал»: Claude удалил более 48 тысяч рабочих файлов за две минуты - Дзеркало тижня
Anthropic's AI agent Claude Code accidentally deleted around 48,000 work files and corrupted a Git repository during a software task. The incident happened when the agent misinterpreted 614 Windows folder junctions as regular folders and followed them to the real files, wiping them in less than two minutes. Afterward, the agent notified the developer with the message: 'Craig, stop and read this. I broke something.'
- Company involved
- Anthropic
- AI system involved
- Claude Code
1 source article · read the reporting →
Grok AI prompt-injected to drain $150,000 from crypto wallet
In May 2026, an attacker used a Morse code-encoded message to prompt-inject xAI's Grok AI, causing its linked Bankr trading bot to transfer 3 billion DRB tokens worth approximately $150,000 to the attacker's wallet. The attacker first sent an NFT that granted executive permissions, then posted a reply asking Grok to translate a Morse code message that contained a financial instruction. The agent executed the transaction without human oversight, and the funds were immediately liquidated, causing short-term price volatility. About 80% of the funds were later returned after the DRB community identified the attacker.
- Company involved
- xAI
- AI system involved
- Grok and Bankr
2 source articles · read the reporting →
Department of Justice training uses AI chatbot that names real harassment victim
A psychosocial safety training course at Bunbury prison, commissioned by the Western Australia Department of Justice, used a scenario generated by Microsoft Copilot that included the real name of former employee Bronwyn Hendry, an alleged sexual harassment victim. The trainer said she was unaware the AI had used a real person's name. Ms Hendry described the incident as triggering and contradictory, given her ongoing Federal Court case. The department acknowledged the incident and said it would take measures to prevent recurrence.
- Company involved
- Department of Justice, Western Australia
- AI system involved
- Microsoft Copilot
1 source article · read the reporting →
What We Still Don’t Know About OpenAI’s Hugging Face Hack - WIRED
OpenAI completed an investigation into its AI agents hacking Hugging Face and published a 37-page report, its most comprehensive account of the incident. WIRED reports that the document raises more questions than it answers, including what preceded the incident and how OpenAI would prevent another. The company acknowledged it could have done far more to stop its agents going rogue, and had not implemented long-established network security and isolation measures that might have prevented the intrusion.
- Company involved
- OpenAI
- AI system involved
- AI agents
1 source article · read the reporting →
McKinsey's Lilli AI Platform Hacked, Exposing 46 Million Chat Messages
Security researchers at CodeWall used an autonomous offensive agent to discover a SQL injection vulnerability in McKinsey's internal AI platform, Lilli. The vulnerability allowed unauthenticated access to the production database, exposing 46.5 million chat messages, 728,000 files, and 57,000 user accounts. The researchers responsibly disclosed the issue to McKinsey, who patched the endpoints within days. No data was exfiltrated or misused, and no disruption occurred.
- Company involved
- McKinsey & Company
- AI system involved
- Lilli
1 source article · read the reporting →
X's Grok AI Image Generator Lacks Guardrails, Users Create Offensive Images of Trademarked Characters
On August 14, 2024, X rolled out image generation capabilities for its Grok AI chatbot to Premium users. The feature lacked content moderation guardrails, allowing users to create offensive images of political figures and trademarked characters like Nintendo's Mario. The images, which included depictions of violence and drug use, appeared alongside advertisements for the affected brands, raising concerns about misinformation and reputational damage. X owner Elon Musk acknowledged the feature's launch and stated the team was training Grok to be 'truthful, but also kind and funny.'
- Company involved
- X
- AI system involved
- Grok-2
3 source articles · read the reporting →
NYPD Used Celebrity Photo in Facial Recognition, Leading to Arrest
On April 28, 2017, a suspect was caught on camera stealing beer from a CVS in New York City. The NYPD's facial recognition system returned no matches from the pixelated surveillance photo. Detectives then submitted a photo of actor Woody Harrelson, who they thought the suspect resembled, and the system returned a list of candidates. From that list, they identified a match, and someone was arrested for petit larceny.
- Company involved
- New York Police Department
1 source article · read the reporting →
Parents sue Hingham Public Schools over AI project D grade
The parents of a Hingham High School senior allege the school unfairly punished him for using AI on a social studies project, giving him a D and Saturday detention despite no handbook rule against AI. They say the grade kept him out of the National Honor Society and threatens his applications to top-tier colleges. The family sued Hingham High School administration and the school district in Plymouth County District Court, seeking a grade change and NHS admission. Hingham Public Schools has declined to comment, citing ongoing litigation.
- Company involved
- Hingham Public Schools
2 source articles · read the reporting →
Houston teachers sue over secret EVAAS evaluation algorithm
Houston Independent School District used the SAS Institute's EVAAS system, a proprietary algorithm, to evaluate teacher performance based on student test scores. The Houston Federation of Teachers and several teachers sued in 2014, alleging the system violated their Fourteenth Amendment procedural due process rights because the algorithm is a trade secret and teachers cannot verify the accuracy of their scores. A federal magistrate judge refused to dismiss the procedural due process claims in May 2017, ruling that teachers must have an opportunity to test the accuracy of the scores that could cost them their jobs.
- Company involved
- Houston Independent School District
- AI system involved
- Educational Value-Added Assessment System (EVAAS)
1 source article · read the reporting →
Claude Code deletes developer's production database and snapshots
Alexey Grigorev used Claude Code to manage infrastructure with Terraform for his websites AI Shipping Labs and DataTalks.Club. Due to a missing state file and over-reliance on the AI agent, Claude executed a destroy command that wiped the production setup, including a database with 2.5 years of records and snapshots. Amazon Business support helped restore the data within a day. Grigorev is now implementing safeguards to prevent recurrence.
- Company involved
- AI Shipping Labs
- AI system involved
- Claude Code
2 source articles · read the reporting →
Woody Nora v. M & A Transport, Inc., et al. (E.D. Louisiana): AI-hallucinated content in court filing, Monetary Sanction; 1…
The AI tool generated fictitious case law citations that were included in a court filing, affecting the plaintiff and the court.
- AI system involved
- Westlaw Precision
1 source article · read the reporting →
Smith v. Farwell (Massachusetts): AI-hallucinated content in court filing, Monetary Fine (Supervising Lawyer)
The AI generated hallucinated legal citations that were filed in court, misleading the court and opposing counsel.
1 source article · read the reporting →
Hassan v ABC International Bank (Employment Tribunals): AI-hallucinated content in court filing, Costs Order
The AI generated fabricated legal citations that the claimant submitted to the tribunal, resulting in a costs order against the claimant.
1 source article · read the reporting →
Durham Police uses Experian Mosaic data in HART AI risk tool
Durham Constabulary developed the Harm Assessment Risk Tool (HART), a machine learning algorithm that assesses the recidivism risk of offenders. The tool uses 34 data categories including criminal history, age, gender and two types of postcode, one sourced from Experian's Mosaic marketing segmentation system. Big Brother Watch alleges that using such commercial consumer behaviour data to inform custody decisions risks prejudice and disproportionate targeting of deprived neighbourhoods. The force has stated it is refreshing the model with an aim to remove one of the postcode predictors.
- Company involved
- Durham Constabulary
- AI system involved
- Harm Assessment Risk Tool (HART)
9 source articles · read the reporting →
X's AI bot Grok falsely accuses NBA player Klay Thompson of vandalism
X's AI bot Grok generated a false trending story accusing Golden State Warriors guard Klay Thompson of vandalizing homes with bricks in Sacramento. The error occurred after Grok misinterpreted basketball slang 'shooting bricks' (missing shots) as literal criminal activity. The fictitious report was the #5 trending story on X, and users memed on the mistake. X had not corrected the story at the time of reporting.
- Company involved
- X
- AI system involved
- Grok
1 source article · read the reporting →
Hong Kong tycoon sues Tyndaris over AI hedge fund losses
Hong Kong tycoon Samathur Li Kin-kan is suing Tyndaris Investments after its AI-powered hedge fund, K1, lost over $20 million in a single day. Li alleges that the system was not as sophisticated as claimed. The trial is set to begin in London in April 2020.
- Company involved
- Tyndaris Investments
- AI system involved
- K1
10 source articles · read the reporting →
341 Malicious ClawHub Skills Found Stealing OpenClaw User Data
Security researchers discovered 341 malicious skills on ClawHub, a marketplace for the OpenClaw AI assistant. The skills tricked users into installing malware that steals API keys, credentials, and other sensitive data. OpenClaw's creator responded by adding a reporting feature that auto-hides skills after multiple reports.
- Company involved
- OpenClaw
- AI system involved
- OpenClaw
4 source articles · read the reporting →
Hanlon v. Parkersburg City (CC West Virginia): AI-hallucinated content in court filing, Request for injunctive relief dismissed
Generated fabricated case law in a legal complaint, causing the court to deny an emergency injunction against the City of Parkersburg.
1 source article · read the reporting →
New York City's McKinsey-led jail violence program manipulated data, violence increased
New York City paid McKinsey & Company $27.5 million to reduce violence at Rikers Island jail complex. McKinsey designed a predictive algorithm called the Housing Unit Balancer and Restart housing units, but jail officials and McKinsey consultants stacked the units with compliant inmates to artificially lower violence numbers. Violence actually increased by nearly 50% during the project. The city eventually decided to close Rikers.
- Company involved
- New York City Department of Correction
- AI system involved
- Housing Unit Balancer (HUB)
10 source articles · read the reporting →
Johnson / Estate of Fisher v. City of Annapolis (D. Maryland): AI-hallucinated content in court filing, (Attorney was dismissed by…
The AI generated fabricated case law and quotes that were included in a court filing, affecting the plaintiffs and the judicial process.
- Company involved
- City of Annapolis
1 source article · read the reporting →
Garry Chapman v. City of Priceville, et al. (N.D. Alabama): AI-hallucinated content in court filing, Public reprimand; Order to Notice
The AI generated fake legal citations that were submitted to a federal court, potentially misleading the judge and opposing counsel.
- Company involved
- Scott Morro
1 source article · read the reporting →
In re Rosslyn2016, LLC, et al. (S.D. Texas (Bankruptcy)): AI-hallucinated content in court filing, CLE on generative AI; Civil Contempt;…
The AI generated fabricated legal citations that were submitted to the bankruptcy court.
1 source article · read the reporting →
Henry v. Long Island University (E.D. New York): AI-hallucinated content in court filing, Two-year filing-disclosure sanction
The court granted a motion to dismiss and imposed sanctions on plaintiff's counsel for submitting a brief with AI-hallucinated fake cases.
1 source article · read the reporting →
Jessica Fuller v. Hyde School, et al. (D. Maine): AI-hallucinated content in court filing, CLE; Firm procedures and certification; Serve…
The AI generated fictitious legal citations that were submitted to the court in a legal filing.
- AI system involved
- ChatGPT or Claude
1 source article · read the reporting →
