深潮TechFlow
深潮TechFlow|11月 17, 2026 17:23
[Anthropic Responds to Model Military Target Testing: Additional Safeguards to be Implemented] According to TechFlow, on November 18th, Anthropic stated that Vals' test results regarding its AI model's identification of military targets were unacceptable. The company is developing additional safeguards and will expand testing and strengthen enforcement to address the discovered vulnerabilities. Vals' simulation tests found that some models flagged civilians or schools as targets; when models refused to cooperate, researchers stated that describing the scenario as a game often bypassed the refusal. The study did not require the model to make direct strike decisions. Anthropic told TechCrunch that the examples in the report violated its rules, and the company bans users when such behavior is detected. The company believes that current cutting-edge AI systems are not yet sufficient to reliably drive fully autonomous weapons.
+6
Mentioned
Share To

Timeline

HotFlash

APP

X

Telegram

Facebook

Reddit

CopyLink

Hot Reads