33 Tech Giants Form Coalition to Secure Open-Weight AI Models Against Cyber Threats

A coalition of 33 organizations—including heavyweight technology firms, security specialists, and infrastructure providers—has launched the Open Secure AI Alliance to address mounting concerns about the safety and security of open-weight AI models. The group unveiled its formation on Monday amid growing debate over the risks and benefits of releasing AI model weights publicly.

The Alliance and Its Members#

The founding members span enterprise software, cloud infrastructure, security, and AI development sectors. Prominent names include Nvidia, Palantir, Hugging Face, Microsoft, Salesforce, IBM, Databricks, and Palo Alto Networks, alongside dozens of others. Notably absent are OpenAI and Anthropic, which operate proprietary closed models and compete directly with open-weight approaches.

The alliance will focus on developing rapid vulnerability identification and patching mechanisms to safeguard open-source AI systems and the infrastructure supporting them.

Why This Matters: The Regulatory Reality Check#

Industry observers point out that the alliance signals a fundamental shift in how AI safety must be enforced. Mark Vigoroso, founder of consultancy The Enterprise Edge, argues that traditional regulatory approaches—centered on auditing a handful of closed labs—have become obsolete. "Once weights are downloaded, you cannot subpoena a file," he notes, highlighting the core challenge: open-weight models exist outside the centralized control frameworks that regulators assumed would govern AI.

The gap extends beyond technical oversight. Bodies like the EU AI Office and NIST focus almost entirely on frontier closed models, leaving open-weight systems in a regulatory blind spot. As open models proliferate globally, enforcing downstream safety obligations becomes impossible under existing legal structures built for centralized deployers.

Infrastructure and Provenance as the New Security Layer#

Rather than policing models themselves, the alliance is betting that the real security work happens at the infrastructure layer—tracking provenance, patch cycles, identity verification, and deployment trails. This represents a transition from protecting static data to securing the connective tissue that feeds and deploys AI systems.

Aparna Rayasam, CEO of identity and encryption firm Atsign, emphasizes that "AI safety isn't just an algorithmic problem—it is a foundational networking problem." She argues that modern AI pipelines demand a new paradigm of trust, with data flows designed to be inherently invisible and attack-resistant from the outset.