Anthropic, the artificial intelligence (AI) firm behind Claude, said it is investigating unauthorized access to Mythos, its powerful new model, which the company worries could be a boon for bad actors. The incident has raised fresh concerns about the security of advanced AI tools, prompting Anthropic to sound the alarm about potential misuse by hackers in large-scale cyberattacks.
Claim 30% Off TipRanks
Trade NVDA with leverageMythos Security Breach Came Through a Vendor System
Bloomberg, which first reported the probe on April 21, said a small group of users gained access to Anthropic’s unreleased Mythos AI model through a computer system normally reserved for its external vendors. According to the report, the individuals involved are members of a private Discord channel dedicated to tracking unreleased AI models.
They exploited multiple access points to gain entry to the restricted system. One method involved credentials belonging to an employee at a contractor supporting Anthropic’s operations. Another involved guessing the model’s URL based on familiarity with Anthropic’s naming conventions for its existing models.
Since gaining access, members have reportedly used Mythos regularly. They later provided Bloomberg with proof in the form of screenshots and a live demonstration. Anthropic confirmed it is investigating the claims and said the access may have come “through one of our third-party vendor environments.”
The company added that it found no evidence that the activity extended beyond the vendor environment used to access systems during model development. AI labs commonly rely on third-party contractors for tasks such as testing, although it remains unclear which vendor was involved in this case.
Mythos Breach Raises Concerns Over Security
The reported unauthorized access has added to concerns surrounding Mythos. It raised questions about whether Anthropic can keep sensitive technology out of the hands of exploiters. After all, the company recently experienced two major leaks in just one week, exposing Claude Code’s sensitive data and information about Mythos before its preview launch.
Anthropic has stated that Mythos significantly outperforms other AI systems in terms of vulnerability detection. Because of its capabilities, the private company withheld the model from general release, granting access to only a closed but growing group of companies and financial institutions. This includes Apple (AAPL), Google (GOOGL), Microsoft (MSFT), Nvidia (NVDA), Goldman Sachs (GS), JPMorgan Chase (JPM), and others, to test the model and strengthen their cybersecurity defenses.
Anthropic has also faced criticism from some observers who argue it may be overstating the model’s capabilities as competition intensifies with rival OpenAI. Still, financial institutions in the U.K. are expected to begin using Mythos soon, while banks in Japan and Canada are in discussions about adoption.
Which AI Stocks Are a Strong Buy?
While Anthropic remains a private company with no publicly traded stock yet, investors interested in AI stocks can explore well-known players such as Nvidia (NVDA), Alphabet Class A (GOOGL), Micron (MU), Microsoft (MSFT), and Meta Platforms (META). All of these stocks have been rated Strong Buys by various Wall Street analysts. To get more information on their performance ratings and price targets, visit TipRanks Stocks Comparison Center.
