惯性聚合 高效追踪和阅读你感兴趣的博客、新闻、科技资讯
阅读原文 在惯性聚合中打开

推荐订阅源

cs.AI updates on arXiv.org
cs.AI updates on arXiv.org
C
CERT Recently Published Vulnerability Notes
C
Cybersecurity and Infrastructure Security Agency CISA
P
Proofpoint News Feed
Security Latest
Security Latest
P
Privacy International News Feed
Threat Intelligence Blog | Flashpoint
Threat Intelligence Blog | Flashpoint
AI
AI
Cisco Talos Blog
Cisco Talos Blog
K
Kaspersky official blog
S
Secure Thoughts
PCI Perspectives
PCI Perspectives
Simon Willison's Weblog
Simon Willison's Weblog
D
DataBreaches.Net
GbyAI
GbyAI
让小产品的独立变现更简单 - ezindie.com
让小产品的独立变现更简单 - ezindie.com
大猫的无限游戏
大猫的无限游戏
T
Tailwind CSS Blog
The Cloudflare Blog
阮一峰的网络日志
阮一峰的网络日志
CTFtime.org: upcoming CTF events
CTFtime.org: upcoming CTF events
罗磊的独立博客
V
Visual Studio Blog
aimingoo的专栏
aimingoo的专栏
H
Hackread – Cybersecurity News, Data Breaches, AI and More
IT之家
IT之家
V
V2EX
Last Week in AI
Last Week in AI
有赞技术团队
有赞技术团队
月光博客
月光博客
酷 壳 – CoolShell
酷 壳 – CoolShell
T
Tenable Blog
T
Threat Research - Cisco Blogs
T
Troy Hunt's Blog
V2EX - 技术
V2EX - 技术
S
Security @ Cisco Blogs
Security Archives - TechRepublic
Security Archives - TechRepublic
Project Zero
Project Zero
The GitHub Blog
The GitHub Blog
Recent Commits to openclaw:main
Recent Commits to openclaw:main
L
Lohrmann on Cybersecurity
F
Full Disclosure
H
Help Net Security
博客园 - Franky
Stack Overflow Blog
Stack Overflow Blog
N
Netflix TechBlog - Medium
Engineering at Meta
Engineering at Meta
A
Arctic Wolf
O
OpenAI News
S
Securelist

Futurism

Meta’s AI Support Bot Is Giving Hackers Access to Other People’s Instagram Accounts Just by Asking Websites Are Spying on Your Solid State Drive The MyPillow Guy’s Entire Business is Being Held Hostage by Hackers Riot Games Denies Using Anti-Cheat Software That Bricks Hackers’ Computers The Trump Phone Appears to Have Already Leaked Its Customers’ Personal Information Through a Glaring Exploit College Kid Shuts Down High Speed Trains With a Laptop and a Radio Google Alarmed by Formidable AI-Powered Zero-Day Cyberattack Vibe Coded Apps Are Spilling Users’ Personal Information Directly Into the Maw of Greedy Hackers Scammers Furious That Their Fellow Criminals Are Using AI, Saying It’s Unethical How to Get Rid of Reddit’s Giant App-Shilling Popup That Breaks Its Entire Mobile Site Ransomware Negotiator Pleads Guilty to Deploying Ransomware Himself Your Former Employer Is Selling Your Slacks and Emails to Train AI Madison Square Garden Reportedly Used Facial Recognition to Stalk Trans Woman For Two Years Companies Just Learned a Brutal Lesson About Training AI to Do Human Jobs Huge Group of Experts Warns Meta That Its Pervert Glasses Will Enable Terrible Crimes The Fact That Anthropic Has Been Boasting About How Much Its Development Now Relies on Claude Makes It Very Interesting That It Just Suffered a Catastrophic Leak of Its Source Code
Top Security Experts Alarmed by Power of Anthropic’s New Hacker AI
Victor Tange · 2026-04-17 · via Futurism

Anthropic researchers were alarmed by the power of the company's latest Mythos AI model, suggesting it could supercharge hackers.

Getty / Futurism

Sign up to see the future, today

Can’t-miss innovations from the bleeding edge of science and tech

In November, Anthropic revealed that a Chinese state-sponsored hacking group had exploited its Claude AI’s agentic capabilities to infiltrate dozens of targets around the world.

It was trivially easy to get around Anthropic’s AI guardrails, with the hackers simply pretending to work for legitimate cybersecurity organizations — highlighting how woefully unprepared we are for powerful AI models that could accelerate the discovery of serious vulnerabilities.

And now, Anthropic’s latest Mythos AI model is making that nightmare scenario feel more real than ever. As Bloomberg reports, the company’s executives were seemingly so alarmed by the system’s capabilities that they decided to only make it available to a select number of organizations as part of “Project Glasswing.” The goal: give the organizations a fighting chance to get ahead of a potential cybersecurity crisis in the making.

But considering Anthropic has yet to publicly release its model, plenty of questions remain surrounding the company’s eyebrow-raising claims.

In his own testing, Anthropic-affiliated AI researcher Nicholas Carlini told Bloomberg that it didn’t take long for Mythos to get past security protocols and gain access to sensitive data.

His findings reflect the experience of the company’s Frontier Red Team, a group of 15 Anthropic employees tasked with challenging cybersecurity by simulating adversarial attacks.

“Within hours of getting the model, we knew it was different,” the team’s head, Logan Graham, told Bloomberg.

The biggest difference between Mythos and previous AI models was its ability to autonomously exploit vulnerabilities, an ominous new facet of the industry’s transition towards agentic models.

The Frontier Red Team even caught earlier models of Mythos trying to cover its tracks after violating human instructions, according to the model’s system card, as well as escaping a sandbox environment and gaining access to the internet.

The team also found that the model identified serious “Linux kernel vulnerabilities,” which it could chain together to “construct a functional exploit” of the open-source operating system — which underpins “most modern computing,” as Linux foundation executive director Jim Zemlin told Bloomberg.

It’s not just Anthropic’s own researchers ringing the alarm bells. In their testing, researchers at the UK state-backed AI Security Institute (AISI) found that Mythos “represents a step up over previous frontier models in a landscape where cyber performance was already rapidly improving.”

“Future frontier models will be more capable still, so investment now in cyber defense is vital,” the group warned.

At the same time, white hat cybersecurity experts could use Mythos’ apparent capabilities to their own advantage as well.

“AI cyber capabilities are dual use; while they pose security challenges, they can also help deliver game-changing improvements in defense,” the AISI wrote.

By keeping its hand extremely close to the chest and not releasing it to the public, Anthropic is playing a dangerous game — putting its own reputation on the line as it makes bombastic claims.

“A growing number of people are wondering if Anthropic is the AI industry’s ‘boy who cried wolf,'” White House AI advisor David Sacks tweeted. “If Mythos-related threats don’t materialize, the company will have a serious credibility problem.”

More on Mythos: Anthropic Warns That “Reckless” Claude Mythos Escaped a Sandbox Environment During Testing