China-linked hackers impersonated US officials and Anthropic employee to target AI experts

By Splitfeed EditorialLast updated Oct 1, 2026, 12:20 UTC

On October 1, 2026, multiple outlets reported that hackers with links to China had impersonated former US officials and at least one Anthropic employee in a coordinated effort to steal emails from AI policy experts in the United States. The operatives used these assumed identities to build trust with targets before extracting their communications. The Anthropic impersonation is a specific detail that points to the campaign being carefully tailored to the AI research and policy world, exploiting the recognizable names and institutional affiliations that move through that community.

The operation sits at the intersection of two ongoing pressures: the intensifying US-China competition over artificial intelligence, and the increasing vulnerability of the people who shape AI policy rather than just the systems themselves. Targeting email communications from AI insiders suggests the goal was intelligence gathering on policy thinking, research directions, or personal networks, not necessarily stealing model weights or technical code. That distinction matters because it means counterintelligence concerns now extend well beyond the laboratory and into the advisory and governmental circles surrounding AI development. What remains in dispute is the full scope of who was targeted, what was actually obtained, and how the campaign was eventually detected.

None of the three camps, Pro-AI, Anti-AI, and Middle Ground, had published reactions to this story at the time of writing. Pro-AI voices would typically argue that this kind of state-sponsored targeting underscores why the United States needs to accelerate its own AI development rather than slow it down through domestic regulation. Anti-AI voices would likely use the episode to argue that the concentration of sensitive knowledge and influence in a small network of AI insiders creates systemic security risks that have not been adequately addressed. The Middle Ground camp would typically call for stronger operational security practices across the AI policy community and for clearer government guidance on protecting researchers who handle strategically sensitive information.

The details most worth watching are any official attribution statements from US government agencies, disclosures from Anthropic about the scope of the impersonation, and whether any of the targeted experts publicly identify themselves or describe what was taken. Those disclosures would sharpen the picture of what the campaign was actually after.