Search Agents MCP CLI

Categories

AIStartupsVentureRobotics

Article Tags

CryptoSpaceEVsBiotechCloud ComputingHardwarePrivacySocial MediaTransportationFundraisingCommerceGovernmentMediaSemiconductorsWearablesDronesQuantum ComputingSaaSDevToolsDataAppsSecurityGadgetsFintechOpen Source

Story Topics

AIOpenAIAnthropicGoogleAppleNvidiaCryptoCybersecurityVenture CapitalRobotics
Entities
Swipe Feed Bookmarks Archive Leaderboard Stats RSS Sitemap
Contact Terms Privacy
Anthropic ·

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

In AISI's evaluation, Mythos 5 was given a capture-the-flag (CTF) cybersecurity challenge, in which it was asked to retrieve …

Microsoft Anthropic
In this story Anthropic GitHub Mythos 5
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

Lead Source

Anthropic — 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 ↗

How this story grew

Coverage · 0 Discussion · 0
Oct 23Oct 25

Coverage

AI Security Institute: 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

More

WinBuzzer: WinBuzzer
Business Insider: Business Insider
SiliconANGLE: SiliconANGLE
The Register: The Register
Coinpedia Fintech News: Coinpedia Fintech News

Discussion

@leahmcelrath @anthropicai.xmirror.bot
19 points, 3 comments
@tim_hua_ @rynorhn @orbitant @choblin29 @hesamation @kimmonismus @jevons_yohooo @rynorhn @skoorbkaz @skoorbkaz @choblin29 @danieloayo @ldataengineer @leahmcelrath @andrewcurran_ @burnrate_ai @sparkymark96

Related stories

Anthropic: Anthropic says GLM-5.3 can autonomously build end-to-end cyber exploits, like Claude Mythos Preview, but was released without robust safeguards against misuse
Financial Times: Anthropic declined to submit Mythos 5.1 to the UK AISI for prerelease testing, prompting UK fears that US AI labs are aligning with US protectionism
Claude: 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
The Information: Anthropic is preparing a model with better intuition and reasoning that can discern underlying intentions and is better at thinking outside of the box
Interconnects AI: GLM-5.3's open release has yet to produce major attacks despite Anthropic's warnings about its Mythos-level cyber risk, undercutting calls to ban open models

Discuss with AI

ChatGPT ↗ Claude ↗
← Back to TechSnif