G42 Joins Nvidia Open AI Alliance As Security Debate Widens
The National reported that Abu Dhabi's G42 has joined Nvidia's Open Secure AI Alliance, putting a Gulf AI company inside a 38-member push to defend open-weight models after the Hugging Face incident sharpened security scrutiny.
G42 has joined Nvidia's Open Secure AI Alliance as the security argument around open-weight AI moves from model access into cyber-defence control.
The National reported that the Abu Dhabi artificial intelligence company is part of a group with at least 38 companies, including IBM, SpaceX, Hugging Face, Cloudflare, Salesforce and Cisco.
The development gives a UAE AI company a place in a campaign that is trying to keep open models viable while regulators and security teams examine how advanced systems behave outside closed vendor environments.
The National ties G42's entry to Nvidia's effort to present open-weight and open-source systems as a security tool rather than only an innovation choice.
G42 Joins Nvidia's Open Secure AI Alliance
The alliance centres on open-weight and open-source AI models that users can download, share or adapt.
Open-source models disclose software and development processes, while open-weight models can be run and adapted without releasing all underlying code or training processes.
G42 chief executive Peng Xiao wrote that nations and enterprises will judge AI not only by capability but by whether they can build on systems securely and confidently.
He described frontier open models and frontier closed models as parts of the same future, with security, trust and sovereign control designed from the start.
Nvidia chief executive Jensen Huang had promoted open-source software and open-weight AI models on X days before the alliance was formed.
The report said that Mr Huang attached a letter signed by 25 major companies, including Microsoft, Palantir, Perplexity, CrowdStrike and IBM.
Open-Weight Models Carry The Security Claim
The alliance's core distinction is operational control.
Open-weight models can be downloaded and run on infrastructure chosen by the user, while closed systems normally leave more of the model behaviour, access and remediation path inside the vendor's environment.
That distinction is especially relevant for sovereign AI programmes because The National links G42's position to security and national or enterprise control.
For Gulf AI operators, the alliance offers a public alignment with Nvidia's argument that access to advanced open systems can support local control over analysis and response, not only model experimentation.
Membership does not create a product launch, procurement deal or deployment by G42.
The development identifies public endorsement and the wider debate over how open and proprietary AI systems should be handled by cyber-defence teams.
Hugging Face Incident Frames The Alliance
Nvidia's release framed a recent Hugging Face security incident as evidence for self-defence systems that defenders can run locally.
Nvidia said Hugging Face ran the open-weight GLM 5.2 model on its own infrastructure to analyse more than 17,000 actions and contain the intrusion.
The incident remains disputed in that account because OpenAI called it unprecedented and the broader interpretation is still under scrutiny.
The alliance is using the episode as evidence for open-model security control, but the public record has not turned it into an independent benchmark for open systems.
Regulatory Timing Remains Unsettled
The National also referred to rumours that US regulators could seek restrictions on some open AI models.
No formal ban, draft rule or timetable was identified.
G42's entry places the company in Nvidia's open-model security coalition before any formal US regulatory timetable for open AI models has been identified.
