White House bans Mythos 5 after Anthropic algorithm breaks NSA defenses in record time
An innovative artificial intelligence algorithm, called Mythos 5 and developed by Anthropic, was quickly withdrawn from circulation by a direct order from the White House. The measure specifically prevents foreign entities from using the tool, on the grounds that the model, with its exceptional ability to locate software flaws, represents a high risk to the national security of the United States. The decision underscores growing concern among governments about the destructive potential of highly advanced AI technologies.
Mythos 5 algorithm penetrates sensitive agency systems in hours
Unpublished and alarming details about the incident were published by the English magazine The Economist, adding gravity to the situation. As revealed by General Joshua Rudd, who took over as director of the National Security Agency (NSA) in March, Mythos 5 managed to access “practically all” the security agency’s confidential systems. Most surprising was the speed of the intrusion, which took just a few hours, contrasting sharply with the weeks that would be expected in conventional security tests. This discovery was communicated to Senator Mark Warner, vice-chairman of the influential US Senate Intelligence Committee, signaling the seriousness with which the episode is being treated at the highest levels of government.
The agility with which Mythos 5 overcame the NSA’s security barriers demonstrates a qualitative leap in artificial intelligence vulnerability analysis capabilities. While human red teams spend weeks or even months identifying weaknesses in complex systems, AI accomplished the task in a fraction of that time. This fact raises urgent questions about the effectiveness of current defenses against cyber threats that can be driven by such powerful algorithms, highlighting the need to reevaluate cybersecurity strategies and develop countermeasures with the same sophistication.
Understand how the NSA used AI in security exercises
Despite the intrusion, it is crucial to clarify that artificial intelligence did not operate autonomously to attack the systems. Initial assessment suggests that the NSA employed the model in a “red team” exercise. This is a standard methodology in which experts simulate cyber attacks to identify and correct vulnerabilities in their own networks, seeking to improve the resilience of defenses. The intention was to test the limits of its security barriers, using a cutting-edge tool to simulate the tactics of sophisticated adversaries. Although controlled, the effectiveness of Mythos 5 in this simulated environment raised a warning about what could happen if the technology fell into the wrong hands or was used maliciously.
Leia Também
Messi reaches 19 goals in FIFA World Cups with great free kick for Argentina in clash with Jordan
Cristiano Ronaldo shines beyond the pitch at the 2026 World Cup with business empire and family luxury
At 41, Cristiano Ronaldo reaffirms his historic legacy and prepares to shine in the 2026 World Cup
Algorithm history shows ability to find faults
In previous evaluations, the Mythos 5 had already demonstrated its impressive effectiveness. In a test authorized by the Mozilla Foundation, the algorithm identified 271 security flaws in the Firefox browser, one of the most used globally. Faced with growing concerns about the dangers inherent in such a robust tool, Anthropic decided to restrict access to the algorithm, making it available only to researchers and institutions that received specific approval, in an attempt to control its use and mitigate potential risks. This limited access policy highlights the company’s responsibility in handling technologies that have such significant impact power.
Fable Version Quick Release and Removal
On June 9, Anthropic released Fable, a supposedly more controlled version of Mythos 5, incorporating safeguards designed to prevent dangerous actions. However, Fable’s availability was extremely brief. Just three days later, on June 12, the White House intervened again, issuing an order that the template be immediately removed from the website. The swift action was triggered by allegations that Amazon engineers had discovered methods to circumvent these protections, which led to warnings from competent authorities about the persistence of risks, demonstrating the difficulty in creating effective barriers against advanced AI capabilities.
The recurrence of the ban, now for the Fable version, demonstrates the urgency and seriousness with which the US government is treating the issue of controlling powerful AIs. The White House’s direct intervention so soon after the release of a “safe” version reflects the perception that, currently, the risks of misuse outweigh the controlled benefits. This episode adds an important chapter to the debate on the governance of artificial intelligence, especially regarding the ability to predict and neutralize vulnerabilities that the technology itself can create.
Anthropic and the debate on the ethical control of artificial intelligence
Anthropic has a notable history of publicly warning about the potential dangers of its own artificial intelligences before making them available, a practice seen with previous versions based on models like GPT-2 and GPT-3. However, Mythos 5 and Fable represent a substantial technological advancement, exhibiting superior capabilities and considerably greater damage potential when compared to their predecessors. This leap in complexity and power raises profound discussions about control and ethics in the development of high-power artificial intelligence. Incidents involving these algorithms intensify the global debate about how to balance technological innovation with safety and responsibility, a challenge that requires collaboration from governments, businesses and the scientific community to establish effective standards and ensure that AI benefits humanity without compromising its safety.
