Skip to content
Submit incident
AII-1000-73173
ReportedReported by fewer than three independent publishers, with no official finding.

Anthropic limits early access to Claude Mythos AI amid vulnerability concerns

Anthropic has delayed public release of its Claude Mythos AI model and restricted early access to a limited set of companies due to concerns it could expose critical vulnerabilities in global infrastructure. The company is negotiating with governments including India on controlled access to identify and mitigate risks before wider launch.

25 April 2026 Β· Anthropic Β· Security and cyberattacks
25 April 2026
AII-1000-73173View source β†—
Quick facts
Category
Security and cyberattacks
Country
United States
Source
Economic Times
Date
25 April 2026
Drafted by
Raidu Incident Detection

What happened

Anthropic has delayed the public release of its Claude Mythos AI model and is restricting early access to a select group of companies to identify and fix critical infrastructure vulnerabilities before wider launch.

Anthropic is in discussions with the US and allied democracies, including India, on securing critical infrastructure ahead of public release. Conversations focus on timing, scope and structure of international access to identify and plug vulnerabilities in global software and cross-border risks.

Reported impact

Claude Mythos is said to be the most powerful AI model developed to date and is expected to expose deep-seated vulnerabilities in the infrastructure of companies globally.

Finance Minister Nirmala Sitharaman convened a meeting with heads of various banks, the Reserve Bank of India and MeitY to understand the potential cybersecurity risks associated with Mythos, with the finance ministry stating the threat was 'unprecedented and requires a very high degree' of attention.

Index analysis

Nasscom stated that as AI systems evolve to autonomously identify and chain vulnerabilities across platforms, the potential for cascading, cross-border risk becomes significantly higher.

Governance controls

Anthropic has given early access to a select group of companies to fix their systems before public release, though no Indian companies were included on this list.

Nasscom wrote to Anthropic requesting inclusion in Project Glasswing and access to Mythos to build cyber resilience for Indian technology firms whose code is used globally. MeitY is also in discussions with Anthropic executives regarding early access for Indian companies.

Classification

Category
Security and cyberattacks
Industry
Digital security
Country
United States
Harm status
Potential harm
Intent
Unintentional
Timing
Pre-deployment
Harm types
Economic or property, Public interest
Harmed
Business
AI systems
Claude Mythos

Sources and evidence

This record was researched and written by the Index, based on the following original reporting.

Economic Times
Before Mythos goes public, Indian IT also wants access - The Economic Times
25 April 2026