Proofpoint: China-aligned TA419 impersonated a senior Anthropic employee and ex-White House officials to phish US AI policy experts
Proofpoint reported on Oct 2, 2026 that TA419, a China-aligned espionage actor, ran credential-phishing campaigns against US AI policy experts, posing as former OSTP principal deputy director Lynne Parker, former State Department chief economist Heidi Crebo-Rediker and (in February) a senior Anthropic employee asking a think-tank analyst for feedback on the military's use of Claude. The aim appears to be intelligence on US AI policy and export controls.
Key facts
- Campaigns: February 2026 (Anthropic impersonation; subject line referenced the debate over US military use of Claude) and July 2026 (Parker, Crebo-Rediker)
- Technique: benign rapport-building emails (e.g. invitation to an 'AI Policy Advisory Committee'), then a shortened URL to a fake OneDrive adversary-in-the-middle page that relays logins to real Microsoft infrastructure and captures passwords, MFA approvals and session cookies (Frameless BitB, Cloudflare Turnstile)
- TA419 has targeted US and Japan think tanks, defense contractors, universities and law firms since at least April 2025 (Proofpoint)
- The Anthropic employee's name was not disclosed
What happened
Proofpoint's threat research team described a China-aligned group using the identities of trusted AI-policy figures, including someone at Anthropic, to get US AI-policy analysts to log in to fake cloud pages.
Why it matters
AI policy, especially export controls and military use of frontier models, has become an espionage target in itself, and lab staff are now useful personas for social engineering.
Changelog
- 2026-10-03: created (morning web run, AI Weekly)
Sources (3)
- officialProofpoint: Hallucinating Credibility, China-aligned TA419 impersonates its way into US AI policy circles
- pressCyberScoop: AI policy circles targeted in China-linked phishing operation
- pressHelp Net Security: China-aligned TA419 phishing AI policy experts
id: 2026-10-02-ta419-impersonates-anthropic-ai-policy-phishing · updated 2026-10-03 · open in the interactive timeline