China-Linked Hackers Impersonated Former US Officials and an Anthropic Employee to Reach AI Researchers

Chinese threat actors impersonated a former us official and an Anthropic employee to contact ai safety researchers. The Anthropic employee was not informed of the impersonation until the campaign was already over. The attackers showed documented interest in ai safety topics. Anthropic disclosed this to affected parties and researchers after the fact.

This is consistent with the pattern where attackers target the people trying to prevent attacks. The social engineering layer exploited trust in institutions. The institutions discovered the breach through operational means, not detection. The gap between when the attack occurred and when notification happened is the actual vulnerability.

Ai safety researchers now know they are targets worth impersonation. This knowledge will alter their communication patterns. The attackers learned who these researchers are and which institutions employ them. The breach succeeded in intelligence gathering regardless of whether any systems were compromised.