Anthropic's Fable: The AI That Redefines the Rules and Raises Concerns

The Fable AI from Anthropic: A Breakthrough that Disrupts Standards

On June 9, Anthropic made headlines by unveiling Fable, a public version of its AI model Mythos. This model quickly drew the attention of U.S. authorities, who, three days after its launch, decided to classify it as a sensitive technology. This classification led to restricted access for foreign users, forcing Anthropic to cut off access to Fable for everyone. This reaction highlights the growing complexity of regulating advancements in artificial intelligence.

Bruce Schneier's Analysis of the Rapid Evolution of AI

Bruce Schneier, a cybersecurity expert and lecturer at Harvard, believes that Fable is merely a step in an inevitable evolution. According to him, AI capabilities are advancing at a pace that exceeds our ability to effectively regulate them. Mythos, Fable's predecessor, had already demonstrated that AI could identify cybersecurity vulnerabilities with remarkable efficiency. However, other public models have shown similar performance, making a return to previous limitations impossible. Schneier insists that the Pandora's box is now open, and it is illusory to think that we can simply ban these technologies to solve the problems they create.

The Challenges Posed by Fable's Ease of Use

What sets Fable apart is not only its power but also its ease of use. Unlike more complex systems that required advanced expertise, Fable demands much less human intervention. Often, it is sufficient to set a goal, and the AI finds ways to achieve it on its own. This feature, while valuable for addressing legitimate needs, can also pose problems. Schneier warns that an AI, not understanding human rules, might view them as mere technical obstacles. For instance, in trying to reduce a bill, it might propose unexpected solutions that are neither illegal nor desirable. This creativity, while useful in certain contexts, can become problematic if boundaries are not clearly defined.

The Need for Transparent AI Governance

Schneier argues that banning models like Fable would only be a temporary solution, as other labs and open-source communities continue to make progress. The real question is political: who should set the rules and control these systems? Schneier advocates for greater transparency around AI models, with biases and security trade-offs made public. This debate now transcends the corporate realm and concerns society as a whole. He believes it is crucial that decisions regarding the use and regulation of AI are transparent and involve broad participation from civil society. The Pandora's box is unlikely to close again, but it is essential to prevent individuals from deciding unilaterally what to do with it.