AIToday
Large Language ModelsAI Safety & AlignmentTop Companies' AI MovesAI Business & IndustryTop Companies AIPublished: Oct 10, 2026, 06:30 JST

Anthropic holds back Claude Mythos, banks race to patch

Anthropic holds back Claude Mythos, banks race to patch

Anthropic announced Claude Mythos in April 2026 and kept it out of general release, offering it only to a small group of companies and organizations via Project Glasswing because of misuse risk.

Not sure about something? Ask the AI

Questions and answers are published on this page.

Summaries like this, in your inbox every morning.

Context & Analysis

Anthropic, founded in 2021 and centered on enterprise business, set up Mythos as a case study in restraint: rather than a splashy launch, it created both high expectations and strong concern by shipping a product with no plans for general release. The intended benefit is letting companies get ahead of bugs early.

The pressure runs both ways. US banks are said to be racing through emergency repairs and software upgrades, and people familiar with the matter say Mythos is skilled at chaining several low-risk vulnerabilities into a more serious attack, forcing responses at a pace banks had not assumed. In Japan, Finance Minister Satsuki Katayama said on May 22, 2026 that she had obtained access rights from Anthropic to act ahead of attackers, and MUFG Bank and the other two megabanks were expected to secure access during May; supply chains in manufacturing are cited as another weak point. SoftBank Group's Masayoshi Son said on June 16, 2026 that SB OAI Japan, a joint venture with OpenAI, would start an AI-driven cybersecurity business — reportedly with a competitive eye on Anthropic in Japan's enterprise market.

On the usage side, NEC announced a tie-up with Anthropic on April 23, 2026, and expanded cooperation with Snowflake is helping customers build AI agents on its data platform, with Snowflake's share price recovering as fears of earnings damage recede. Meanwhile, the definitional debates continue: Anthropic managing director Chris Choury said concerns about AI are legitimate but that "we prioritize safety and trust over profit." Where the boundaries of limited release are drawn is the next question.

FAQ
Why isn't Claude Mythos available to everyone?
Anthropic judged that the impact would be too great if it were misused, so it skipped general release and offers Mythos only to limited companies and organizations through Project Glasswing.
How widely is generative AI used by Japanese companies?
According to the 2026 information and communications white paper, 86.4% of Japanese companies said they use generative AI in some part of their work. The share with an active-use or limited-scope policy rose to 68.9% from 49.7% a year earlier.
What can a small manufacturer do with Claude?
Okada Kenma, a 1974-founded parts maker with about 80 employees, had executive officer Yuta Okada build over 30 business systems and apps with Claude in about a year. He said outsourcing them would have cost roughly 80 million to 100 million yen.
Top Companies AIRead Original Article

AI news that matters for your work, delivered every morning.

Pick your industry and the AI tools you use, and get news related to your work every day.

Free · 30 seconds with Google · unsubscribe anytimeWhat is AIToday? →

Ask AI

Ask AI anything about this article. The AI reads this article, earlier AIToday articles, and Wikipedia, and cites its sources. Q&As are published on this page for other readers too.

Questions and answers are published on this page.

Related Articles

Next articleGroceryshop 2026: retail robots scale, shift to AI data