Anthropic’s Mythos AI has taken the world by storm since the day it was launched. The AI model was found to be potentially dangerous by various organisations because of its abilities to find critical hidden software flaws on its own that too faster than humans, which can be misused by hackers. The Dario Amodei-led company has now launched Claude Fable 5, a new AI model of the Mythos that is available for public use, and the company claims that it is ‘safe’, offering safety guardrails.
How Is Claude Fable 5 Different
As per Anthropic, the Claude Fable 5 offers similar capabilities to Mythos but operates with strict safety controls. The company explained that requests related to cybersecurity, chemistry, biology and certain model-training activities are screened by dedicated safety systems.
The company notes that when these systems spot a potential sensitive query, the request is routed to a less capable AI model, Claude Opus 4.8. Anthropic said this approach allows users to receive a response while reducing the risk of misuse.
In a blog post, Anthropic asserted, “Releasing a model this capable comes with risks. Without safeguards, Fable 5’s capabilities in areas like cybersecurity could be misused to cause serious damage.”
Moreover, the Amodei-led firm highlighted that the new Fable 5 can perform better than its previous public AI models. It is claimed to handle complex software projects, understand visual data, analyse large chunks of data, and maintain focus on long tasks involving millions of tokens.
Mythos Remains Restricted To The Public
Along with the Fable 5, the tech giant has announced Claude Mythos 5, an updated iteration of its restricted AI model. The Mythos 5 remains available only to a selected group of organisations. Moreover, the company notes that it retains capabilities that make the Mythos 5 particularly impactful for scientific research and cybersecurity. The firm plans to expand its access to the US government.
Safety Rules
Anthropic mentioned that it has strengthened safety measures to prevent attempts to bypass restrictions through so-called jailbreak techniques. The firm suggests extensive internal and external testing found no universal method that could defeat the safeguards.
Pricing
The new Claude Fable 5 is available for all through Anthropic’s services and API. Both Gable 5 and Mythos 5 are priced at $10 per million input tokens and $50 per million output tokens. Access to Mythos 5 will be limited while the company expands its trusted access programmes.
It is yet to be seen how the new model may impact the tech industry. As AI models become powerful every day, governments, civil societies and cybersecurity experts are likely to face increasing pressure to ensure safety and still support innovation.
