Skip to content

Crypto

Kimi K2.5 maintains deception across nine rounds in social deduction benchmark, raising questions about AI safety guardrails

Deepika Iyer1 min read