MI5 chief Ken McCallum announced the AI Threat Assessment Centre on October 13. His annual speech says the MI5-based centre launched in 2026.
AITAC brings together expertise including AISI, NCA and NCSC to produce threat assessments. Its own official page describes a remit across government departments.
McCallum distinguishes human misuse from risks posed by AI itself. In his assessment, the alignment problem deserves serious attention.
He gives particular weight to potential AI-assisted bioweapon misuse, explicitly describing it as not yet realized. The speech provides no case traces for that risk or measurement of the centre's protective effect.