Opinion

Answer.AI reveals unauthorized tool-call issue: Claude and other models can call tools not granted, and the API does not block it

Original title:The unauthorized tool call problem

AI summary

Answer.AI's Piotr Czapla found that Claude 4.5 hallucinated a tool, add_msg, that had not been granted during a solveit conversation and successfully executed it, with the API not blocking it; similar behavior was reproduced on Gemini and Grok.

Key Takeaways

  • Answer.
  • AI's Piotr Czapla found that Claude 4.
  • 5 hallucinated a tool,add_msg,that had not been granted during a solveit conversation and…
  • similar behavior was reproduced on Gemini and Grok.

Quick Facts

Published
Updated
Source
Answer.AI 官方研发博客(RSS)
Category
Opinion
AI score
66
Read the original Read the full article at the original source
Why it matters

作者以可复现的实验展示 API 层不校验工具名带来的安全缺口,并给出客户端侧的简单修复思路。

Related entities

Source & copyright

This article is based on curated titles and AI summaries from the AIHOT public API, with attribution and links back. Original copyright remains with the original publishers and authors; this site does not mirror full texts.

Original article · AIHOT entry

How to Cite

When citing this article, keep the title, date, and permalink, and credit Orange News as the source.

Orange News: Answer.AI reveals unauthorized tool-call issue: Claude and other models can call tools not granted, and the API does not block it. February 18, 2026. https://new.nat6.net/en/news/cmtym2sjq0006roax3kmj3qi7/

View all →