Mythos, Anthropic AI, discovers over 2,000 unknown software vulnerabilities in seven weeks

In a recent test that lasted only seven weeks, Mythos artificial intelligence, created by Anthropic, found over 2,000 software vulnerabilities unknown by the industry. The achievement represents a significant leap in the ability of AI agents to apply governance on a scale, reinforcing cybersecurity at a time when digital attacks become more frequent and sophisticated.

How does Mythos encounter so many unknown software vulnerabilities?

Unlike traditional automated scanners, Mythos operates as an AI agent that understands the context of each piece of code. It analyzes entire repositories in search of patterns that indicate security flaws, without relying on known signatures. This allows her to discover gaps that have never been catalogued — the so-called "zero-day" or unknown vulnerability. The tool also isolates access to confidential data, ensuring that sensitive information is not exposed or compromised during scanning.

What is the impact of these software vulnerabilities on the security market?

With the [AI arms race accelerating](/?noticias/corrida-armamentista-de-ia-acelera-com-novos-modelos-da-openai-deepseek-e-anthropic), the volume of code produced by developers grows exponentially. Manual audit methods do not follow this scale. Mythos fills this gap by offering autonomous screening, reducing the time between the introduction of a failure and its correction. For companies dealing with sensitive data, such as banks and healthtechs, this can prevent catastrophic violations.

What did Anthropic reveal about the software vulnerabilities discovered?

According to the company, more than 2,000 vulnerabilities were identified in controlled environments, simulating real production scenarios. Anthropic did not disclose specific details about which software was tested, but stated that Mythos was able to cover both open source and owner. The news was originally published by Fox News, which highlighted the potential of the tool to transform the way security teams perform audits.

Is Mythos accessible only to large security companies?

Although the ad mentions business applications, Anthropic plans to make the technology available to independent developers and startups, expanding access to a proactive defense. The company also discusses partnerships with regulatory bodies, such as the Bank of England, which already [assess the impact of Mythos on the financial sector](/?noticias/setor-financeiro-do-reino-unido-preparado-para-mythos-e-ias-de-fronteira-afirma-banco-da-inglaterra).

Frequently Asked Questions about software vulnerabilities with AI

Mythos replaces human auditors?

Nope. It functions as a complementary layer, automating the initial scan for experts to focus on critical corrections.

What kind of vulnerabilities does she encounter?

From memory leaks and SQL injection to logical failures in access permissions.

Is the tool commercially available?

Anthropic has not yet announced a public release date, but has already conducted tests with selected partners.