Anthropic, a leading company in artificial intelligence development, announced on April 9, 2026, that it would limit the release of its new model, Mythos, due to its advanced capabilities in identifying vulnerabilities in widely used software.
The company has decided not to release the model to the general public, opting instead to make it available exclusively to a select group of major corporations and organizations managing critical online infrastructure, such as Amazon Web Services and JPMorgan Chase.
This selective distribution strategy has sparked intense debates within the technology sector. Anthropic justifies the measure as a way to allow these companies to prepare against potential cyberattacks by anticipating threats that Mythos could reveal.
However, there are questions about other interests behind the decision. As reported by TechCrunch, OpenAI is considering a similar approach for its next digital security tool, suggesting a trend among AI industry giants.
Cybersecurity market experts have also expressed their opinions on the real impact of Mythos. Dan Lahav, CEO of Orca Security, emphasized that while AI tools’ identification of flaws is relevant, the potential exploitation of a vulnerability depends on various factors, such as the possibility of combining it with other breaches.
Lahav raised doubts about the actual severity of the findings made by Anthropic’s model, questioning whether they represent significant risks in isolation or in attack chains.
Meanwhile, a cybersecurity startup called Aislesec, which develops AI-based solutions, claimed to have replicated much of the capabilities attributed to Mythos using smaller, open-source models. Representatives from Aislesec argued that the results reinforce the idea that there is no single definitive model for cybersecurity, but rather that performance depends on the specifics of each task or threat faced.
Another point of discussion is the economic impact of this restrictive approach. David Crawshaw, a software engineer and CEO of the startup exe.dev, pointed out that the selective release of Mythos could create a cycle of contractual dependency between Anthropic and large companies.
This strategy would make it difficult for competitors to use distillation techniques, a process that allows new AI models to be trained from existing systems in a more accessible manner. With limited access, Anthropic could consolidate its market position, keeping distillation companies at a competitive disadvantage.
The clash between labs developing cutting-edge models, like Anthropic, and companies betting on open-source solutions is redefining the dynamics of the artificial intelligence sector. Restrictions on access to technologies like Mythos are seen by some analysts as a way to protect the commercial interests of major developers, highlighting their corporate products in an increasingly competitive market.
It remains unclear whether Mythos poses a concrete threat to internet security or if its capabilities have been overestimated. Anthropic has not commented on speculations that the decision to limit the release is also linked to concerns about the distillation of its models. Meanwhile, the technology community continues to monitor the developments of this controversy, which could influence future release strategies in the AI sector.
Original published at O Cafezinho.