"If a software dependency contains malicious code, we have mature practices for discovering it, tracking its provenance, and reducing its impact. AI models are different. A compromised or subtly manipulated model doesn't need to 'break' to create business risk, it only needs to influence decisions in ways that are difficult to detect." semgrep.dev/blog/2026/ai-suppl

And to demonstrate, one of the researchers prompted their way into a backdoor in an open weight model: theregister.com/ai-and-ml/2026

This is kind of an incredible website:

Why your stuff stopped lasting, who profited, and what's still built right.

worseonpurpose.com

Show more
Librem Social

Librem Social is an opt-in public network. Messages are shared under Creative Commons BY-SA 4.0 license terms. Policy.

Stay safe. Please abide by our code of conduct.

(Source code)

image/svg+xml Librem Chat image/svg+xml