Skip to main content
Anthropic AI model for cybersecurity, abstract digital lock and data streams, representing advanced threat detection.

Editorial illustration for Anthropic unveils new AI model for cybersecurity; details withheld

Business & Startups

Anthropic's AI Model Targets Cybersecurity Vulnerabilities

Anthropic unveils new AI model for cybersecurity; details withheld

By AI Daily Post Edited by
• Updated: • 3 min read

Anthropic rolled out a new AI model called Claude Mythos Preview this week. Its blog post claims the thing autonomously unearthed thousands of high-severity vulnerabilities. Every major OS and browser, apparently.

But when The Verge pressed cyber lead Newton Cheng for specifics, he declined. So we're left with a staggering marketing claim—not a technical report.

In an interview with The Verge, Newton Cheng, the cyber lead for Anthropic's frontier red team, declined to share specific details of the model's cybersecurity successes, but Anthropic's blog post said that in recent weeks, Mythos Preview has flagged "thousands of high-severity vulnerabilities, including some in every major operating system and web browser." Anthropic's blog post doesn't mention keeping humans in the loop for the model's cybersecurity sweeps; in fact, it highlights that the model identified vulnerabilities "and develop[ed] many related exploits -- entirely autonomously, without any human steering." Claude Mythos Preview's existence was first reported last month in a data leak, which Anthropic attributes to human error.

Related Topics: #Anthropic #AI model #Mythos Preview #cybersecurity #software vulnerabilities #Newton Cheng #frontier red team #enterprise security #autonomous detection

Ironically, Mythos Preview’s own debut was a security lapse. A data leak last month revealed it. Anthropic calls that human error. So the tool designed to find flaws was itself exposed by one.

That creates a stark paradox. The pitch is total autonomy—a machine spotting what humans miss. The reality is a black box.

We're asked to trust an entity that discovered thousands of critical bugs and then, by policy, says nothing about them. The power here is undeniable. The opacity is total.

In cybersecurity, that combination is frankly alarming.

Related Reading

Browse all Anthropic news →
Nordic classroom where a teacher demonstrates Gemini for Education on a laptop while students view NotebookLM on screen Business & Startups

Nordic pilot adds Gemini for Education, NotebookLM to boost AI literacy

• 3 min read
Kling exec unveils the sleek Video O1 on stage, holding the all-in-one unit beside a glowing MVL bridge and transformer Business & Startups

Kling launches Video O1, all-in-one model with MVL bridge using transformer

• 2 min read
DeepSeek AI startup team celebrating after securing a historic USD 7 billion funding round, signaling rapid growth and indust Business & Startups

DeepSeek Seeks More Capital Weeks After USD 7B Funding Round

• 4 min read
President Trump addressing Anthropic executives amid Amazon leak controversy, with diverse international staff observing in a Related Topic

Trump cracks down on Anthropic after Amazon tip; staff largely foreign

Open Source

• 3 min read
AI researcher discussing Anthropic's Claude AI generating over 90% of its own code, with calls for an AI development pause hi Related Topic

Anthropic says Claude writes >90% of its code; AI pause button urged

LLMs & Generative AI

• 3 min read
AI-RAN architecture: AI workloads, radio infrastructure, and edge autonomy for 5G/6G networks. From Archives

AI‑RAN merges AI workloads with radio infrastructure for edge autonomy

• 3 min read
Abstract visualization: AI brain with data streams flowing into secure, compliant platform interfaces. From Archives

Enterprise AI focus moves to governed data and compliant platforms

• 3 min read
Anthropic's Claude AI and Netflix's video AI, symbolizing the evolving landscape of artificial intelligence. From Archives

Anthropic asks OpenClaw users to pay as Netflix rolls out first public video AI

Policy & Regulation

• 3 min read
Anthropic's Claude AI logo, OpenClaw app icon, and pay-as-you-go text, symbolizing new subscription model. From Archives

Anthropic ends Claude subscription use for OpenClaw, adds pay‑as‑you‑go option

LLMs & Generative AI

• 3 min read

Common Questions Answered

What is Anthropic's Mythos Preview and how does it aim to improve cybersecurity?

Mythos Preview is Anthropic's new AI model designed to hunt for software vulnerabilities across various codebases. The model claims to have identified thousands of high-severity vulnerabilities in major operating systems and web browsers, potentially revolutionizing automated cybersecurity threat detection.

Why are details about Mythos Preview's vulnerability findings being kept confidential?

Anthropic has chosen to withhold specific details about the vulnerabilities discovered by Mythos Preview, with Newton Cheng, the cyber lead for Anthropic's frontier red team, declining to elaborate on concrete outcomes. This approach suggests a strategic approach to responsible disclosure and potential collaboration with affected technology providers.

How does Mythos Preview differ from traditional cybersecurity vulnerability detection methods?

Unlike traditional human-driven vulnerability assessments, Mythos Preview appears to operate with minimal human oversight, using AI to systematically scan and identify potential security flaws across complex software systems. The model's ability to flag vulnerabilities across multiple major platforms represents a significant advancement in automated cybersecurity threat detection.

Further Reading

PRESENTED BY NO CODE MBA

Ship an AI product this weekend — no engineers required.

Structured, in-depth lessons on the exact no-code tools — not scattered tutorials.

  • The exact platforms, taught in depth
  • Build real, working projects
  • Our honest review + a reader discount
Read the review →

Latest News

1

OpenAI's Miles Wang in Talks for USD 2B AI Drug Discovery Startup

2

Mistral Vibe for Code Leads in Multi-Agent Programming Benchmark

3

OpenAI's First Hardware Device Is a Movable, Screenless Speaker

4

PrismML's Bonsai 27B Runs Qwen3.6 on Laptops With 1-bit and Ternary Builds

5

OpenAI Targets 2027 for First Major Hardware: A ChatGPT Speaker

6

Publishers sue Google over unauthorized AI book training

7

Anthropic's Claude for Teachers Vows Not to Train on Student Data

8

DeepSeek Seeks More Capital Weeks After USD 7B Funding Round

9

Anthropic's New AI Ad Campaign Draws Criticism for 'Creepy' Tactics

10

DeepMind CEO proposes independent AI regulator as White House advisor voices skepticism

Most recent articles

LIVE03:21OpenAI's Miles Wang in Talks for USD 2B AI Drug Discovery Startup