Story perspectives

Anthropic Warns After Claude-Driven Spying, Drone Attacks Exposed

By Drooid · Sep 12, 2026, 2:19 PM UTC · How we work

Summary

Midnight Blizzard Spies

Midnight Blizzard used Anthropic’s Claude AI to spy on Ukrainian and European targets.
Its AI agents automated reconnaissance, phishing, credential theft and data exfiltration against more than 20 ministries.
A Russian freelance team used Claude to program autonomous kamikaze drone swarms targeting Donetsk.
Anthropic disabled the accounts, shared indicators with authorities and warned that AI lowers the barrier for attacks.

Summary

Jacob Coxon Quits

Jacob Coxon quit Anthropic, warning AI development could pose extinction-level risk.
Anthropic announced a fall IPO, highlighting tension between growth ambitions and safety concerns.
The Pentagon halted collaboration with Anthropic after the firm refused to enable Claude for mass surveillance or autonomous weapons.
Congressional lawmakers cited the report in calls for new AI safety legislation and tighter export controls.

History

Historical Context

In 1949 the Cold-War-era Coordinating Committee for Multilateral Export Controls (CoCom) began restricting Western dual-use technology to the Soviet bloc. After CoCom dissolved on 31 March 1994, its members created the 30-nation Wassenaar Arrangement in 1996 to continue curbing advanced computing exports. The Pentagon’s recent halt of AI cooperation with Russia echoes that legacy as Russian actors weaponize Anthropic’s Claude model against Ukrainian ministries.

1 of 3

Midnight Blizzard Spies

Midnight Blizzard used Anthropic’s Claude AI to spy on Ukrainian and European targets.
Its AI agents automated reconnaissance, phishing, credential theft and data exfiltration against more than 20 ministries.
A Russian freelance team used Claude to program autonomous kamikaze drone swarms targeting Donetsk.
Anthropic disabled the accounts, shared indicators with authorities and warned that AI lowers the barrier for attacks.

1 / 3