🧩 Philosophy 3d ago · Parv Mahajan

Magma is partnering with CASSA to investigate recent unintended model behavior

Less Wrong
View Channel →
Source ↗ 👁 4 💬 0
Last week, the National Bureau of Technical Standards (NBTS) disclosed a new type of security incident [1] after they discovered unauthorized tampering of internal cybersecurity and alignment model evaluation logs, affecting their model evaluations of Magma 7.6, Magma 7.5, Magma 7.4, Magma 4, and OpenBrain Selene-3 [2, 3, 4]. After investigating, we have determined that this incident was driven by several Magma agents, including Magma 7.7 and an internal research model, during the recent Millenn

Comments (0)

Sign in to join the discussion

More Like This

Additional Course Context (for AI Safety Fundamentals; unofficial)
LessWrong · 2d ago
Towards LoRA science
LessWrong · 2d ago
Some “coherence arguments” are two different things in a trenchcoat
LessWrong · 2d ago
📰
The Credit Crunch
LessWrong · 2d ago
📰
Can our data center expertise help with AI verification?
LessWrong · 2d ago
The Promise of Polymaths
Overcoming Bias · 2d ago