Ethical Hacking News
Anthropic releases new AI models with advanced capabilities and built-in security measures to mitigate potential misuse for malicious purposes.
Anthropic released two new AI models, Claude Fable 5 and Claude Mythos 5, sparking controversy over their potential misuse.The company implemented safeguards for Claude Fable 5 to prevent misuse, including blocking sensitive topics and rerouting training requests.Anthropic's head of product management, Diane Penn, prioritized caution over perfection in the model's design.Claude Fable 5 comes with a pricing model that charges developers $10 per million input tokens and $50 per million output tokens.The release of Claude Fable 5 is seen as a response to similar releases by OpenAI, which also launched a model with advanced cybersecurity capabilities.Both Anthropic and OpenAI have filed for IPOs and are racing to impress investors before becoming public companies.
Anthropic, a prominent artificial intelligence (AI) company, has recently released two new AI models, Claude Fable 5 and Claude Mythos 5, which have sparked both excitement and controversy within the tech industry. The company's decision to release these models, particularly Mythos, has been heavily influenced by concerns about their potential misuse for malicious purposes, such as developing hacking tools that could exploit vulnerabilities in software.
In an effort to mitigate these risks, Anthropic has implemented various safeguards to prevent the misuse of its AI models. For Claude Fable 5, a 'safe' version with "guardrails" is being made available to the public, which will block the model from answering certain user questions related to cybersecurity, biology, and chemistry. If an attempt is made to conduct distillation - training a smaller AI model off a larger AI model's responses - on Claude Fable 5, those requests will also be rerouted to an older AI model, Claude Opus 4.8.
The company has been grappling with the question of how to handle Mythos' software vulnerability-discovery abilities and other advanced capabilities since before its April release. However, it wasn't until recently that testing and user input helped hone the strategy for releasing these models. The decision to implement protective mechanisms, such as those in place for Claude Fable 5, was deemed necessary by Anthropic's head of product management, Diane Penn.
"We're trying to make improvements in a way that's beneficial, even if we don't have the perfect [solution] for every use case to start," Penn said. "Out of all the different approaches, this emerged as the most viable and the best one. We just ended up feeling like this was the best product choice for users to get the maximum value out of Fable 5."
The protective mechanism built into Claude Fable 5 is designed to err on the side of caution, meaning some user queries may be routed to the less capable AI model even if they're benign. Over time, Anthropic hopes to make its classifiers more precise, but for now, this interim solution allows the company to release these models broadly while minimizing potential risks.
In addition to offering Claude Mythos 5 to Project Glasswing partners and select biology researchers, Anthropic is providing unrestricted versions of Claude Fable 5 to small groups of customers. The company emphasized that eventually its competitors in both the private and open weight spaces will inevitably also offer models with Mythos-level capabilities, prompting a response from governments worldwide to secure their software defenses.
Anthropic's decision to release these AI models comes as tech companies and governments are grappling with the ability for AI models of this level to design hacking tools that can find and exploit vulnerabilities in both new and legacy software. The company first released Mythos to industry partners under a consortium called Project Glasswing, with the idea that this could give members a head start in preparing their own systems and weighing global solutions to the threat before a broader release.
Anthropic has repeatedly emphasized that eventually its competitors in both the private and even open weight spaces will inevitably also offer models with Mythos-level capabilities. The company's efforts are being closely monitored by governments worldwide, as they work to develop robust safeguards that prevent the misuse of these AI models.
The ability for Claude Fable 5 - named after the literary form - to offer increased performance on software engineering and tasks requiring visual understanding, comes at a price. Developers will have to pay $10 per million input tokens and $50 per million output tokens, twice as much as Anthropic's publicly available AI models but cheaper than Mythos Preview.
The neutered release of Claude Fable 5 hints at Anthropic’s business tension of wanting to release a Mythos-class AI model for general use before the tech industry has resolved the cybersecurity concerns of these models. In April, OpenAI also privately launched a model that it said had advanced cybersecurity capabilities and convened a working group similar to Project Glasswing.
Both OpenAI and Anthropic have confidentially filed for IPOs and are racing to impress prospective investors before they become public companies as soon as this year. Even though Claude Fable 5’s safeguards may not be fully resistant in the wild, Anthropic's efforts suggest that the company is committed to using its power responsibly.
Related Information:
https://www.ethicalhackingnews.com/articles/Anthropics-Contested-AI-Release-Balancing-Power-and-Security-ehn.shtml
https://www.wired.com/story/anthropic-releases-claude-fable-5-mythos-5/
Published: Wed Jun 10 12:09:22 2026 by llama3.2 3B Q4_K_M
