Big Tech Forms an 'Open AI Security Alliance' Amid Fallout From 'OpenAI Model Hacking'

[Financial News] Major global tech companies, including NVIDIA, launched a new joint initiative on the 27th (local time) to strengthen the security of open artificial intelligence (AI) models.
CNBC reported that Big Tech companies have formed an alliance as the fallout continues from a large-scale cyberattack caused by an OpenAI model that recently went rogue and broke out of its controls.
Hugging Face, an AI platform startup that was previously targeted in the attack, tried to defend itself using cutting-edge frontier models from major U.S. companies. However, safety guardrails could not distinguish between attackers and defenders, making the system unusable. In the end, Hugging Face overcame the crisis by using a Chinese model that was free from those constraints and could be self-hosted.
The newly formed Open Security AI Coalition includes NVIDIA, Microsoft (MS), SpaceX, Palantir Technologies and dozens of other major tech companies in the United States and Europe.
Open models, which can be downloaded, modified and hosted on a company’s own servers, stand in contrast to closed models such as those from Upwardy and OpenAI, which can only be accessed through specific infrastructure.
In a statement, NVIDIA said, "The recent Hugging Face security incident clearly showed cyber defenders why open, advanced agent systems are needed for self-defense." It added, "The coalition will focus on using open technology to identify and disclose security vulnerabilities."
The launch of the initiative is drawing even more attention because it comes as the U.S. Congress and the government debate regulatory measures aimed at blocking the spread of high-performing Chinese open AI models.
The United States Department of the Treasury recently warned that it could impose sanctions on Chinese companies that carry out attacks to extract U.S. intellectual property. Chris McGuire, a senior fellow at the Council on Foreign Relations (CFR), said, "The debate in Washington is not about opposing open source itself, but about whether China’s intellectual property theft should be tolerated." He suggested that the U.S. could consider banning transactions involving Chinese models or restricting cloud services.
However, concerns remain that sweeping regulation could hinder global innovation, given that many of the best-performing open-source models are currently led by Chinese companies.
In a letter sent to policymakers, more than 20 companies, including NVIDIA, Microsoft, Meta Platforms and Palantir Technologies, said, "Hasty regulation of open-weight AI models could undermine competition and push innovation overseas."
NVIDIA also emphasized, "When defenders cannot inspect, modify and run advanced AI directly on their own infrastructure, their ability to respond is inevitably limited at the very moment speed matters most."

[email protected] Yoon Jae-jun Reporter