Anthropic’s Misuse Report: Seven Labs, 190 Million Exchanges, and the Collapse of the Offensive Labor Gap

An analysis of Anthropic's misuse report, published by forkast.news, describes two separate campaigns outlined in that report. In the first, a Russian-speaking actor used prompt injection to compromise an AI vendor's evaluation sandbox and steal production API keys, which were then used to attack approximately 30 AI companies over a period of four days.

The second account concerns affiliates of ShinyHunters, who compromised a SaaS provider and gained access to more than 2,100 Azure AD token sets and 40 corporate tenants within 34 hours. The report notes that 'AI agents performed nearly all the work.' The article's title references seven labs and 190 million exchanges, though the excerpt does not attribute those figures to either account.

The title frames the material around 'the Collapse of the Offensive Labor Gap,' setting the two campaigns alongside the report's observation about agent work. The excerpt states that the Russian-speaking actor's activity spanned four days and that the ShinyHunters affiliates reached the 40 tenants in 34 hours, but it does not name the AI vendor whose evaluation sandbox was compromised, the SaaS provider, the approximately 30 AI companies, or the corporate tenants involved. It also does not specify how the prompt injection was delivered, what was taken beyond the production API keys and token sets, or which organisations produced the seven-lab figure. The two accounts are presented as distinct incidents within the same report.