NVIDIA announced a new initiative called Open Secure AI Alliance, which aims to expand the use of open artificial intelligence technologies in the field of cybersecurity. The alliance, which consists of a total of 27 founding members, includes Microsoft, Dell Technologies, SpaceX, Linux Foundation and NVIDIA, as well as many technology companies operating in different areas of the industry. The main purpose of the initiative is to create a common working ground for faster detection, elimination and responsible sharing of security vulnerabilities by using open technologies. NVIDIA states that this initiative is based on the work of the Akrites initiative within the Linux Foundation and the OpenSSF community.
As artificial intelligence systems become increasingly complex, detecting and analyzing cyber attacks requires new methods. According to the alliance, in order to develop effective defenses, security researchers need to have access not only to commercial and closed artificial intelligence models, but also to open artificial intelligence models. The organization argues that open models offer more flexible use, especially in incident response and digital forensics processes. It is stated that this approach both expands the field of action of security teams and can accelerate the development of new defense tools.
Open Secure AI Alliance focuses on open AI ecosystem
To support this approach, NVIDIA cites a recent security incident involving Hugging Face as an example. According to the information provided by the company, Hugging Face asked commercial closed artificial intelligence models to analyze the attack in the first stage. However, these requests were rejected due to security measures and the necessary forensic analysis could not be carried out. Thereupon, the company turned to the open-weighted GLM 5.2 model, which runs on its own infrastructure. More than 17 thousand transactions were analyzed using the open model and the attack was limited. The alliance argues that this example shows that open models can give security teams more control.
In addition, the founding members plan to contribute in different areas. NVIDIA announced that it will share open artificial intelligence models, model weights, datasets and new agent development tools with the community. HPE, on the other hand, will work on standards and methods that will enable cryptographic verification of services with artificial intelligence agents. Hugging Face will bring the Safetensor format, which enables the safe storage of artificial intelligence model weights, to the initiative. Other participants such as Microsoft, IBM, Red Hat and SpaceXAI also announced that they will contribute to the project with open source artificial intelligence technologies.
The alliance is not limited to technical cooperation only. The group argues that public institutions should also support joint open artificial intelligence infrastructure investments together with companies. However, it calls for regulatory bodies to consider open artificial intelligence models as technologies that can be used for defensive purposes rather than as a risk factor. The statement includes the view that comprehensive restrictions on open artificial intelligence systems may weaken defense capacity and cause advanced artificial intelligence technologies to be concentrated only in the hands of a limited number of closed platform providers.
On the other hand, the announcement of the initiative coincides with a period when regulatory discussions in the field of artificial intelligence are gaining momentum. Various recent news reports suggest that the US administration is considering comprehensive restrictions on China-based open source artificial intelligence models. Open models, which are gaining increasing attention especially due to their low costs, stand out as an alternative to closed systems offered by companies such as OpenAI and Anthropic. Despite this, it is noteworthy that large commercial artificial intelligence developers such as OpenAI, Anthropic, Meta and Google are not among the founding members of the Open Secure AI Alliance.
Open Secure AI Alliance offers a new collaboration model that aims to bring the open source approach more strongly to the field of cyber security. While the flexibility provided by open artificial intelligence models in terms of security research is one of the main pillars of the initiative, the contributions of different technology companies in standards, tools and open technologies also support this goal. The joint projects that the alliance will develop in the coming period and the relationships that will be established with regulatory institutions will be among the main factors that determine how the initiative will be responded to in the industry.
Join Channel