Anthropic opens its most powerful AI model to the public with new safety safeguards

Dubai: Anthropic has released the most powerful artificial intelligence model in its history to the public, a significant step further intensifying race among AI developers to deliver increasingly capable systems while addressing mounting concerns over safety and misuse.

The San Francisco-based company on Tuesday unveiled Claude Fable 5, the first publicly available model from its advanced Mythos class of AI systems. The launch comes just one week after Anthropic disclosed that it had confidentially filed paperwork for an initial public offering, underscoring growing investor interest in the rapidly expanding AI sector.

Anthropic said Fable 5 is its strongest model to date, designed to excel at software engineering, complex research tasks, image analysis and advanced reasoning. The company described it as a substantial leap beyond previous Claude models, while emphasising that it has built extensive safeguards into the system to prevent misuse in sensitive areas.

The Mythos family first emerged in April when Anthropic introduced Claude Mythos Preview, a restricted model that demonstrated an unusual ability to identify vulnerabilities across major operating systems and web browsers despite not being specifically trained for cybersecurity.

The model's capabilities prompted concerns that highly advanced AI systems could potentially be used to discover weaknesses in critical infrastructure, including banking systems, power grids and widely used software platforms.

In response, Anthropic launched Project Glasswing, a security-focused collaboration involving major technology and financial firms, including Amazon Web Services, Apple, Google, Cisco, Microsoft and JPMorgan Chase.

Participating organisations were granted early access to Mythos Preview to help identify vulnerabilities and strengthen their systems before more advanced versions of the technology became available.

Unlike its predecessor, Fable 5 is being released to all users. Anthropic said the decision was made possible by safeguards that automatically block the model from responding to queries in what it classifies as high-risk domains, including cybersecurity and certain biological research topics.

Instead, such requests are redirected to Claude Opus 4.8, a less powerful model that the company believes poses lower security risks.

Anthropic said it subjected the system to extensive internal and external testing to determine whether users could bypass its protections. More than 1,000 hours of independent red-team testing and bug bounty exercises failed to uncover a universal method of circumventing the safeguards, according to the company.

The launch also expands access to the unrestricted Claude Mythos 5 model, which will be made available to approved organisations already participating in Project Glasswing and other vetted programmes. The network now includes around 200 organisations across more than 15 countries.

Anthropic said access to both Fable 5 and Mythos 5 will be priced at $10 per million input tokens and $50 per million output tokens, making them among the company's most expensive offerings. Tokens are the units used to measure AI model usage.

The launch comes as competition intensifies across the AI industry, where companies are racing to develop more capable systems while seeking to reassure regulators, customers and investors that powerful models can be deployed safely.