WorkThe story, in brief
A fundamental flaw leaves LLMs strikingly vulnerable to attack
Researchers just proved LLMs have an unfixable security hole. Here's what that means for your AI deployments.
The KeyNews take
Why it matters
A peer-reviewed finding that LLM vulnerability is foundational—not a bug to patch—forces practitioners to rethink threat models and enterprises to recalibrate risk tolerance for AI-in-production systems.
The key facts
4 to knowPaper presented at ICML 2026 (top-tier conference)
Claims fundamental flaw makes full LLM security impossible
Implications for AI safety and security posture across enterprises
Research challenges the premise that vulnerabilities are solvable through hardening
Go to the source
MIT Technology Reviewtechnologyreview.com
Publisher excerpt: It is impossible to make large language models fully secure against hacks because of a fundamental flaw in how they work, a team of researchers argue in a paper presented at the International Conference on Machine Learning, a top AI conference, this month. The claim has huge implications for the…