As cited
Copy frozen at (site build).
ai security
Anthropic’s new budget model gets much better at ignoring hidden commands
Anthropic released Claude Haiku 5.5, a budget model optimized for speed, with improved capability to identify vulnerabilities and develop exploits compared to its predecessor. The company implemented stricter cybersecurity safeguards on Haiku 5.5 than on Haiku 4.5, though less restrictive than those on more advanced Claude models.
Why it matters: Security teams deploying or interacting with budget large language models need to understand that cheaper, faster models now possess meaningful offensive cybersecurity capabilities, which requires careful evaluation of use cases and access controls.
- Source published
- First seen by Cybersecurity Tracker