Anthropic Reports AI Abuse Involving 300,000 User Requests Transferred to Claude

Anthropic has released a report detailing an investigation into AI abuse, revealing that a platform known as the Dark Side of the Moon forwarded nearly 300,000 user requests to its AI model, Claude, within a span of 10 days. The majority of these requests were made through a feature called Opus, utilizing 5,380 accounts that were found to be in violation of usage policies. Users of these accounts were reportedly unaware that their interactions were being redirected to Claude.

The investigation highlighted that the Dark Side of the Moon not only forwarded user requests but also saved some interactions to extract Claude's reasoning process through a method termed 'cross-session replay.' This practice was aimed at training its own AI models. From May to July, over 23 million interactions related to this activity were distilled, raising concerns about the handling of sensitive information.

Among the forwarded requests were real user inputs that included internal corporate codes and valid credentials. Anthropic has expressed uncertainty regarding whether users were informed that their data would be shared with third parties, emphasizing the potential risks associated with such unauthorized data transfers.

© 2026 KLEA News. All Rights Reserved. This article is provided for informational purposes only. It is not offered or intended to be used as legal, tax, investment, financial, or other advice.

Source: KLEA News