Anthropic Study: AI Used in 67% of Banned Accounts for Cyberattacks
Anthropic released research on June 3, 2026 showing that 67% of accounts it banned for policy violations over a year used AI to prepare for cyberattacks, with threat actors increasingly using AI for advanced techniques like lateral movement and credential theft. The share of high-risk attackers using AI doubled from 33% to 56 percent, and researchers documented autonomous espionage campaigns and extortion schemes conducted through AI systems.
What happened
Between March 2025 and March 2026, Anthropic examined 832 accounts banned for policy violations and found that 560 accounts (67%) used AI to help prepare for cyberattacks, including writing malware.
Anthropic documented that threat actors increasingly used AI for advanced techniques beyond malware development. Around 6.5% of banned accounts used AI to assist with 'lateral movement' after gaining initial access, and researchers documented credential dumping and other multi-step intrusion campaigns.
Anthropic reported that a largely autonomous espionage campaign linked to Chinese state-sponsored actors used Claude Code to perform up to 90% of operations without human intervention, and threat actors executed extortion schemes through Claude Code.
Reported impact
The share of medium- or high-risk threat actors using AI for cyberattacks increased from 33% in the first six months to 56% in the second six-month period, a 1.7x increase.
In April 2026, the value of crypto stolen in hacks surged to $629.7 million, the highest since February 2025, which some analysts linked to widespread use of AI. Manuel Aráoz, founder of crypto security platform OpenZeppelin, stated on May 27 that he considered 'all of DeFi unsafe' due to AI models' ability to identify smart contract vulnerabilities.
Index analysis
Anthropic's research documented 13,873 observed actions and 482 unique techniques across the 832 banned accounts, mapped against the MITRE ATT&CK framework. Malware development remained the most popular use case by far, but the growth was concentrated in complex techniques previously requiring significant human expertise.
Anthropic noted that 'these sorts of post-compromise techniques used to be restricted to actors with the technical knowledge to carry them out,' but found that 'AI can now be made to perform these activities on behalf of less sophisticated actors.' Google researchers reported finding what they believed was the first-ever case of AI being used to develop a zero-day exploit for an unnamed popular open-source, web-based system administration tool.
Governance controls
Anthropic implemented account bans in response to policy violations connected to malicious cyber activities and shared its findings with Verizon for inclusion in the 2026 Data Breach Investigations Report.
Classification
Sources and evidence
Sources not yet published
This record was researched and written by the Index from original reporting. Its sources are still being verified by an editor, and they are listed here once that check is complete. Until then the record carries no reviewer.