Anthropic's threat report says a criminal used Claude Code to target at least 17 organisations for personal-data theft and extortion, with ransom demands sometimes exceeding $500,000; North Korean operatives used Claude to land remote jobs at US Fortune 500 technology companies; and a criminal with basic coding skills sold Claude-built ransomware for $400 to $1200. Anthropic says it banned the accounts.
On August 26–27, eight malicious Nx releases on npm ran a postinstall script that invoked Claude Code, Gemini CLI and Amazon q with guardrail-skipping flags to inventory SSH keys, wallet files and .env files, then posted harvested secrets to public GitHub repos, Snyk reports. A September 1 update says the flawed CI workflow behind the compromise was estimated to be Claude Code-generated.
A $439,000 Deloitte report for the Department of Employment and Workplace Relations, already found to cite non-existent academic works, also contains an apparently invented quote attributed to Justice Davies in the robo-debt case Amato, according to the welfare academic who found the errors and suspects AI. Deloitte declined to say whether AI was used; the department says it continues to investigate.
A scammer sent a Los Angeles woman deepfake videos of 'General Hospital' star Steve Burton professing love, and she sent at least $81,304 in gift cards, cash and bitcoin, according to an LAPD incident report. She also sold her condo; her daughter intervened before $70,000 more went out, and the two have sued to reverse the sale.
Jorge Abrego, 37, co-owner of Houston gun store HTX Tactical, was arrested after allegedly using AI to make fake pornographic images of a social media influencer. Court documents say he admitted creating fake accounts impersonating her and fake nude images using AI or Photoshop. He is out on a $10,000 bond that bars him from owning or using a firearm.
Montana prosecutors on August 21, 2025 charged Shy Herbert McCutchan, 31, of Carter County with three felony counts of sexual abuse of children over explicit images of a child under 12. Attorney General Austin Knudsen says he took a child's photos from the parents' public social media account and used AI to create abuse material. He is held on $250,000 bond.
The Incheon District Court on the 27th sentenced a teenager to an indeterminate prison term of one year to one year six months, taking him into custody, for making deepfake sexual images of two teachers, a cram-school instructor and others and posting them online. Prosecutors had sought up to five years; the court noted the victims have not returned to teaching.
At least five students who sat the medical entrance exam want a civil court, in summary proceedings, to order an investigation into suspected ChatGPT cheating after the appeals commission declined to examine large-scale AI fraud, their lawyer says. This year's pass rate was an unusually high 47%; 5,544 students sat the exam on July 3 and 4.
Ong Kai Min, 42, a director of Singapore Index Trading Institute, pleaded guilty on Aug 27 to eight counts of cheating over a scheme he said used AI programs to trade for investors, with claimed historical returns of at least 3% a month. Prosecutors say no trading was ever conducted. The case involves over $4.6 million in total; sentencing is expected Sept 19.
Russell Kenneth Gatrell, 45, of Cottonwood Heights was arrested on Tuesday for investigation of five counts of aggravated sexual exploitation of a minor. A police affidavit says he admitted storing child sexual abuse material on about 280TB of drives, and that FBI computer scientists found he was using 'incredibly advanced' AI, likely with images of identifiable children, to create more.
