Certitude gets access to Anthropic’s Mythos AI model for offensive security testing
The cybersecurity specialist Certitude Consulting is among the first companies in Europe to receive Red Team Access as part of Anthropic’s Cyber Verification Program. Companies in German-speaking countries can now have their software analysed for security vulnerabilities using an advanced AI model for security research.
For companies in the DACH region, this opens up a new opportunity to have the security of their software and IT systems tested using current AI technology: Certitude can use Mythos’s capabilities as part of professional security assessments, authorised penetration tests and targeted security research projects.
Claude Mythos: capabilities and access
With Claude Mythos, the US-based AI company Anthropic has developed a model with particularly strong capabilities in complex software and security analyses. Mythos can independently analyse extensive codebases, identify previously unknown vulnerabilities and investigate complex interrelationships in software and IT systems.
The model’s capabilities are not only relevant for defence. The same technologies could also be used by attackers to detect and exploit vulnerabilities.
For this reason, Anthropic initially made a conscious decision not to release Mythos to the general public. Access was initially limited to a very small circle of selected organisations as part of Project Glasswing. In April 2026, access was restricted to approximately 200 organisations, including selected technology companies and operators of critical infrastructure.
Anthropic is thus pursuing a controlled approach: the new AI capabilities are initially intended to be available to those who use them to make software and digital infrastructure more secure.
What does Red Team Access mean?
With the Cyber Verification Program, Anthropic has created a special access programme for verified cybersecurity experts and companies.
Defense Access is the first tier of the programme; it allows the use of advanced models for defensive purposes and is aimed at corporate security teams.
Red Team Access is much more strictly limited to organisations whose primary focus is offensive security and security research. This tier requires verification of the organisation and its field of activity by Anthropic, including through publications in this sector. It allows authorised security testing, including:
- Advanced penetration tests and red team assessments
- Identification and analysis of previously unknown vulnerabilities
- Simulation of realistic attack scenarios
- Investigation of complex software and system architectures
- Development and validation of proofs of concept and exploits in an authorised testing environment
This specialised access authorisation takes into account that professional security researchers sometimes require the same technical skills for their work as potential attackers.
Benefits for companies in German-speaking countries
For companies, Certitude’s access to Mythos means they can have their own software, applications and IT systems analysed with the support of an advanced AI model for security research.
Mythos can analyse large volumes of source code, detect potential vulnerabilities and assist security researchers in investigating complex bugs. The use of such models opens up new possibilities, particularly in areas where traditional automated testing methods reach their limits.
The combination of artificial intelligence and human expertise is crucial here. The AI must be guided by security researchers, and the results must be technically validated by experienced experts and evaluated in terms of their actual impact.
“In addition, a technical control and monitoring framework must be developed and operated that embeds the AI in a controlled environment, enabling but also limiting its actions,”
explains Marc Nimmerrichter.
Experience in international security research
Certitude already brings extensive experience in independent security research. Organisations such as Microsoft, Google and the German Federal Office for Information Security (BSI) have already publicly acknowledged the findings of Certitude’s security research specialists.
A recent example is the discovery of two serious vulnerabilities in the Google Chrome browser and in Arm GPU drivers by Certitude security researcher Florian Schweitzer. The vulnerabilities (CVE-2026-9034 and CVE-2026-93374) affect the GPU software stack and, under certain conditions, can enable a so-called sandbox escape.
The vulnerabilities were responsibly reported to the affected vendors, who have addressed them. Google and Arm publicly credit the Certitude researcher as the discoverer of the respective security vulnerabilities.
This research highlights what matters most in modern cybersecurity: not only examining known attack patterns, but also identifying previously overlooked vulnerabilities in complex software and hardware architectures.
With access to Mythos, Certitude can expand this research to include a new technological dimension.
Consequences for software security
The development of powerful AI models presents a new challenge for companies. As AI becomes increasingly capable of identifying security vulnerabilities that were previously difficult to detect, defence strategies must also evolve accordingly.
Certitude aims to provide companies in the DACH region with access to these new opportunities through a combination of advanced AI technology, years of security research experience and in-depth technical expertise.
About Certitude Consulting
Certitude Consulting is a Vienna-based consulting firm specialising in cybersecurity that helps organisations identify digital risks early on and manage them effectively. By closely integrating research, regulation and practical experience, Certitude develops solutions that are not only standards-compliant but, above all, practical and suitable for everyday use.