A post by @unusual_whales on X.
BREAKING: Top AI models regularly misidentify civilians and civilian infrastructure as military targets in tests, per Bloomberg.
@unusual_whales on X
Not investment advice. Muhabbit scores how much a story may matter, not which way to trade.
Leading AI models from Anthropic and other top developers regularly misclassified civilians and civilian infrastructure as military targets during testing, according to a Bloomberg report cited by @unusual_whales. The finding raises accuracy concerns as AI systems are increasingly deployed in defense and decision-support roles. The report did not disclose the number of models tested or the error rate. It comes amid a broader push by AI companies into government and military contracts. What to watch next: whether Anthropic and peers publish evaluation details or adjust training and guardrails for targeting tasks.
Muhabbit's AI wrote this summary of @unusual_whales's post in its own words. It can get details wrong, so read the post on X.
Named in this story
Anthropic, Bloomberg