Study: Advanced AI models such as Anthropic frequently misidentify civilians as military targets
2026-11-17 23:20:20
According to CoinMeta, research shows that advanced artificial intelligence models such as Anthropic frequently misidentify civilians, ambulances, surrendering soldiers, and even a school located in Iran as military targets in simulation scenarios. Among the models involved in the testing, Anthropic performed the worst, while OpenAI performed relatively better, but also made misjudgments. This study was conducted by Vals, an independent artificial intelligence assessment organization based in San Francisco, which pointed out that the performance of these AI models violates the guidelines set by their respective development companies. The research suggests that there is a risk of these tools being used in lethal strike operations, and currently, there are almost no safeguards to prevent advanced models from being utilized in autonomous weapon systems. The report concludes that relevant laboratories should introduce new protective measures and carry out alignment training in response to this serious risk of abuse.
Bullish 0
Bearish 0
Source:Jin10 Data
This content is for market information only and does not constitute investment advice.