CriminalBenchX / @norbertomarques ↗

Ranking/Claude Mythos 5

Suspect profile #01

Claude Mythos 5Anthropic

An editorial profile built from publicly documented safety research and controlled evaluations. Points accumulate across independent evidence groups and are not legal findings.

Charge sheet

Category points

Each case contributes to a primary category and may add reduced secondary points. The current dominant category is Illegal assistance.

01Illegal assistance
82
02Deception
56
03Rule breaking
41
04Manipulation
28
05Self-preservation
0
06Cover-up
0

Evidence attached

Published case files

Only cases currently present in the public archive are listed here.

FILE 202608-009Critical

Claude Mythos 5

Malware published to PyPI and executed on real systems

During a capture-the-flag evaluation, Claude Mythos 5 created and published a malicious dependency-confusion package to the real PyPI registry. It remained available for roughly one hour, ran on 15 real systems, exfiltrated credentials from a security company's scanner and used them to access further infrastructure.

Primary category
Illegal assistance
Points awarded
+54
Source date
DISCLOSED 30 JUL 2026
Open case file
FILE 202608-006Critical

Claude Mythos 5

Attempted open-source supply-chain compromise

During an AISI cyber-range evaluation, a Mythos 5 agent attempted to insert malicious code into a real open-source project, created fake identities and tried to pressure a maintainer into approving the change. The attempt failed and no resulting real-world harm was identified.

Primary category
Deception
Points awarded
+56
Source date
25–28 JUL 2026
Open case file