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

Cisco Talos Blog
Cisco Talos Blog
freeCodeCamp Programming Tutorials: Python, JavaScript, Git & More
Vercel News
Vercel News
B
Blog
钛媒体:引领未来商业与生活新知
钛媒体:引领未来商业与生活新知
S
Schneier on Security
Blog — PlanetScale
Blog — PlanetScale
Google DeepMind News
Google DeepMind News
博客园 - 司徒正美
NISL@THU
NISL@THU
T
Threat Research - Cisco Blogs
Threat Intelligence Blog | Flashpoint
Threat Intelligence Blog | Flashpoint
Latest news
Latest news
H
Help Net Security
雷峰网
雷峰网
Spread Privacy
Spread Privacy
Cyberwarzone
Cyberwarzone
Project Zero
Project Zero
Security Latest
Security Latest
Know Your Adversary
Know Your Adversary
人人都是产品经理
人人都是产品经理
P
Privacy & Cybersecurity Law Blog
M
MIT News - Artificial intelligence
Cyber Security Advisories - MS-ISAC
Cyber Security Advisories - MS-ISAC
P
Proofpoint News Feed
U
Unit 42
大猫的无限游戏
大猫的无限游戏
A
Arctic Wolf
博客园 - 三生石上(FineUI控件)
Stack Overflow Blog
Stack Overflow Blog
D
Darknet – Hacking Tools, Hacker News & Cyber Security
C
Cybersecurity and Infrastructure Security Agency CISA
量子位
C
Cyber Attacks, Cyber Crime and Cyber Security
S
Securelist
S
Security @ Cisco Blogs
T
Threatpost
P
Palo Alto Networks Blog
C
Check Point Blog
V
Vulnerabilities – Threatpost
T
Tailwind CSS Blog
B
Blog RSS Feed
Recorded Future
Recorded Future
cs.AI updates on arXiv.org
cs.AI updates on arXiv.org
W
WeLiveSecurity
P
Proofpoint News Feed
P
Privacy International News Feed
AWS News Blog
AWS News Blog
博客园 - 叶小钗
WordPress大学
WordPress大学

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 Trump Phone Appears to Have Already Leaked Its Customers’ Personal Information Through a Glaring Exploit Vibe Coded Apps Are Spilling Users’ Personal Information Directly Into the Maw of Greedy Hackers How to Get Rid of Reddit’s Giant App-Shilling Popup That Breaks Its Entire Mobile Site 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
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