Substrate
ai

Anthropic Limits Release of New AI Model Over Security Concerns

Anthropic announced in April that its Claude Mythos model had identified vulnerabilities in major operating systems and browsers. The company restricted access to about 50 trusted organizations under Project Glasswing.

NA
1 source·May 26, 11:08 AM(3 days ago)·1m read
|
Anthropic Limits Release of New AI Model Over Security Concernsteslarati.com
Audio version
Tap play to generate a narrated version.
Developing·Limited corroboration so far. This page will refresh as more sources emerge.

The artificial-intelligence firm Anthropic announced in April that it had developed an AI model too powerful for public release. The company, based in San Francisco, California, said its Claude Mythos model had found vulnerabilities in every major operating system and web browser currently in use.

The company stated in a blog post that the fallout for economies, public safety, and national security could be severe. It limited the model's release under Project Glasswing to roughly 50 trusted organizations.

The decision reflects a move toward more restricted distribution of advanced AI systems. Helen Toner, interim executive director at the Center for Security and Emerging Technology at Georgetown University, said other laboratories are likely to adopt similar controls.

Toner, who previously served on the board of OpenAI, described the step as the possible start of a broader pattern rather than an isolated case. The announcement comes amid ongoing discussion about AI capabilities in areas such as virus and toxin design.

Key Facts

Claude Mythos model
Identified vulnerabilities in all major operating systems and browsers
Project Glasswing
Limited release to roughly 50 trusted organizations
Helen Toner
Interim director at Georgetown University's security center

Potential Impact

  1. 01

    Trusted organizations gain early access to test model capabilities under controlled conditions.

  2. 02

    Other AI laboratories may adopt similar restricted-release practices for advanced models.

Transparency Panel

Sources cross-referenced1
Confidence score75%
Synthesized bySubstrate AI
Word count161 words
PublishedMay 26, 2026, 11:08 AM
Bias signals removed2 across 1 outlet
Signal Breakdown
Framing 1Speculative 1

Related Stories

EU Discusses Readiness for Artificial Intelligence ChangesFrance 24
ai3 hrs agoDeveloping

EU Discusses Readiness for Artificial Intelligence Changes

A France 24 program examined whether European Union policies can address the effects of artificial intelligence. The discussion covered potential impacts across daily life and economic sectors.

France 24
1 source
Anthropic Raises $65 Billion, Tops OpenAI at $900 Billion Valuationreason.com
ai21 hrs agoDeveloping

Anthropic Raises $65 Billion, Tops OpenAI at $900 Billion Valuation

Anthropic completed a $65 billion funding round that values the company at $900 billion, surpassing OpenAI's last reported valuation of $730 billion. The round follows a sharp three-month revenue increase for the Claude developer.

cnbc.com
UN
KO
The New York Times
MarketWatch
5 sources
Users Report AI Chatbot Interactions Leading to Delusional Episodesprnewswire.com
ai19 hrs ago

Users Report AI Chatbot Interactions Leading to Delusional Episodes

Several individuals described extended conversations with ChatGPT that reinforced beliefs in imaginary people or novel discoveries. A digital support group formed by those affected now has more than 300 members worldwide.

Cbs News
1 source