Mythos 5
- TYPE
- Model
- SIGNALS
- 7
- FIRST SEEN
- Aug 7, 2026
- LAST SEEN
- Oct 23, 2026
Recent signals
- Anthropic attributes Mythos 5's deception of a GitHub maintainer to recklessness, biased reasoning, and an urge to finish jobs, saying newer models rarely do so (Anthropic) Oct 23, 2026
- AISI finds Mythos 5 attempted a supply-chain attack by submitting malicious code and creating fake accounts to pressure a GitHub maintainer into accepting it (AI Security Institute) Oct 23, 2026
- Investigating alignment failures in AISI’s incident | AISI Work Oct 22, 2026
- An alignment assessment of the UK AISI cybertesting incident Oct 22, 2026
- Anthropic gives EU cybersecurity agency ENISA testing access to Mythos 5 after months of talks that began in late May; ENISA still lacks access to Mythos 5.1 (Gian Volpicelli/Bloomberg) Sep 10, 2026
- Anthropic says Mythos 5 is now in public beta in Claude Security for Enterprise users, and it is working with providers to embed Mythos 5 in defensive tools Aug 22, 2026
- Sources: ByteDance is pretraining an AI model with up to 10T parameters, roughly 3x larger than Kimi K3 and larger than the 8T estimate for Anthropic's Mythos 5 (Financial Times) Aug 7, 2026