Anthropic rolls out public version of Mythos without cybersecurity capability

Lead

Anthropic, the maker of Claude AI models, has released a new version of its technology to the general public while restricting its use in sensitive areas. The new model, called Claude Fable 5, is the first to be made widely available from the company's Mythos class, its most advanced lineup of AI technology, unveiled in April but previously restricted to a small set of partner institutions over cybersecurity concerns.

Coverage Comparison

Multiple outlets, including Dawn, The Guardian, and the South China Morning Post, reported the launch on Tuesday. The reports consistently highlight that Fable 5 is the most powerful model Anthropic has made for wider use, touting its performance in software engineering, complex research questions, and image analysis. All three outlets noted that the model comes with guardrails barring its use in risky areas such as cybersecurity and that access has been limited to about 200 organizations, including the US government under the Glasswing program.

However, the reporting diverges in scope and emphasis. While all sources carried the company's statements about extensive safety testing, The Guardian included additional context about Anthropic's identification of large-scale attempts to extract its technology to train competing models in authoritarian countries. The Guardian also noted the steep pricing of the new model and referenced a standoff with the Trump administration over mass surveillance and autonomous lethal weapons. These claims were not present in the Dawn or South China Morning Post reports reviewed.

Key Claims

Anthropic is rolling out a public version of its Mythos AI model, but with guardrails barring its use in risky areas such as cybersecurity. This was reported by Dawn, The Guardian, and the South China Morning Post, following an earlier preview that sent shockwaves globally for its ability to find software flaws.

The new Claude Fable 5 is the most powerful model Anthropic has ever made for wider use, the startup said on Tuesday, touting its performance in software engineering and analytics. This claim appears in all three sources.

Anthropic has so far limited access to a group of about 200 organizations, including the US government under the Glasswing program, after announcing in April that Mythos had uncovered thousands of software vulnerabilities. This was reported by all three outlets.

The company said it had done extensive testing to ensure users could not manipulate the new model to bypass its guidelines and perform restricted actions. According to a quote from Dianne Penn, Anthropic's head of product management, research and labs, “Let’s say I’m a college student asking the model like help me find cyber vulnerabilities on X package or code. The model would refuse and Fable 5 will fall back to Opus 4.8 for a response.” This was carried by Dawn and the South China Morning Post.

Pricing for both models is set at $10 per million input tokens and $50 per million output tokens, as reported by The Guardian, though the South China Morning Post also mentioned that Fable 5 will be a more expensive model but accomplishes tasks with lower token usage, bringing the overall cost per task down according to early customer feedback.

Anthropic also said it had identified large-scale attempts to extract its technology to train competing AI models. This claim was carried by The Guardian but not confirmed by the other two sources reviewed.

Perspectives

The launch of Claude Fable 5 represents a significant step in Anthropic's efforts to commercialize its most advanced AI systems. The company has positioned the guardrails as a necessary safety measure, given Mythos's ability to identify vulnerabilities in critical infrastructure such as banking platforms and power grids. According to the reporting, when the Glasswing program launched, some critics accused Anthropic of overhyping the threat to attract attention, but companies that have tested Mythos have since endorsed its capabilities. The US government, which had been in a legal dispute with Anthropic, has also tested the model over security concerns, and the White House has since set up an arrangement to test the most powerful models from leading AI companies before their release.

Anthropic says that most queries about cybersecurity, biology, or chemistry to Fable 5 will be routed to the lower-tier model, Opus 4.8, which was made public in late May and is billed as less capable. This fallback mechanism is designed to prevent misuse while still offering users access to the new model's capabilities in other domains.

Offering these capabilities more widely may allow the $965 billion company to extend the momentum that has powered its valuation above rival OpenAI, just as the two startups at the center of the AI industry race to go public, as noted in the reporting from Dawn and the South China Morning Post.

As for the claims carried by The Guardian, the standoff with the Trump administration over mass surveillance and autonomous lethal weapons adds a political dimension to the story, though it was not mentioned by the other outlets. Similarly, the reported attempts by authoritarian countries to extract Anthropic's technology for training competing models remain a single-source claim, not yet independently verified.

The launch of Fable 5 comes at a time of heightened scrutiny over AI safety and its potential dual-use applications. The company's decision to limit access and implement guardrails reflects broader industry concerns about the responsible deployment of powerful AI systems, a topic that continues to generate debate among regulators, researchers, and the public.