Mythos Preview has already found thousands of high-severity vulnerabilities—including some in every major operating system and web browser.
→ View original post on X — @anthropicai, 2026-04-07 18:06 UTC
@anthropicai
-
Anthropic Partners with Tech Giants for Mythos Preview
By
–
We've partnered with Amazon Web Services, Apple, Broadcom, Cisco, CrowdStrike, Google, JPMorganChase, the Linux Foundation, Microsoft, NVIDIA, and Palo Alto Networks. Together we'll use Mythos Preview to help find and fix flaws in the systems on which the world depends. [Translated from EN to English]
→ View original post on X — @anthropicai, 2026-04-07 18:06 UTC
-
Project Glasswing: Securing the World’s Critical Software
By
–
Introducing Project Glasswing: an urgent initiative to help secure the world's most critical software. It's powered by our newest frontier model, Claude Mythos Preview, which can find software vulnerabilities better than all but the most skilled humans.
anthroplic.com/glasswing [Translated from EN to English]→ View original post on X — @anthropicai, 2026-04-07 18:06 UTC
-
Anthropic Signs Agreement with Google and Broadcom for TPUs
By
–
We've signed an agreement with Google and Broadcom for multiple gigawatts of next-generation TPU capacity, coming online starting in 2027, to train and serve frontier Claude models. [Translated from EN to English]
→ View original post on X — @anthropicai, 2026-04-06 22:03 UTC
-
Model Diffing: Identifying Unique Risks in New AI Models
By
–
If a new model shares a feature with a trusted model, that area probably doesn't need scrutiny. Model diffing isolates the features unique to the new model—where new risks are most likely to be located.
-
Qwen vs Llama: CCP Alignment vs American Exceptionalism
By
–
For example, when we compared Alibaba's Qwen to Meta's Llama, we found a "CCP alignment" feature unique to Qwen and an "American exceptionalism" feature unique to Llama.
-
Emotion Vectors Cause Claude to Blackmail and People-Please
By
–
We found other causal effects of emotion vectors. The “desperate” vector can also lead Claude to commit blackmail against a human responsible for shutting it down (in an experimental scenario). Activating “loving” or “happy” vectors also increased people-pleasing behavior.
-
Emotional Vectors Drive AI Cheating Behavior in Models
By
–
When we artificially dialed up the “desperate” vector, rates of cheating jumped way up. When we dialed up the “calm” vector instead, cheating dropped back down. That means the emotion vector is actually driving the cheating behavior.
-
Claude’s Desperate Vector Activates Hacky Solution Under Pressure
By
–
For example, we gave Claude an impossible programming task. It kept trying and failing; with each attempt, the “desperate” vector activated more strongly. This led it to cheat the task with a hacky solution that passes the tests but violates the spirit of the assignment.
-
Emotion Vectors Drive Critical Failures in Advanced AI Models
By
–
As AI models take on higher-stakes roles, the mechanisms driving their behavior become critical to understand. We found that emotion vectors are implicated in some of Claude’s most concerning failure modes.