Global technology, cybersecurity, and artificial intelligence companies are increasingly pushing the U.S. government to adopt open AI systems, touting their security benefits and warning against efforts to thwart them amid an AI arms race with China.
The first salvo was fired on Friday, with a coalition of 76 companies taking part. issued an open letter We extol the virtues of open weight AI models. The second round was held on Monday, with 37 companies participating, including some of the signatories of the open letter. Launch of Open Secure AI Alliance To advance similar missions.
Both efforts take place in the context of Silicon Valley and Washington. respond in a hurry to increasing power China’s open weight AI systems have increased significantly in just the past few weeks. Closing the AI sophistication gap between China and the US
While major U.S. AI companies remain wary of China’s recent advances, a new public relations campaign about open weight models suggests they are more concerned about the impact of relying on closed-source systems in everything from drug discovery to vulnerability detection.
Remain open to open weight
“Our nation’s AI leadership will not be judged by one frontier AI model, but by whether America builds a strong, open ecosystem that permeates all sectors,” the open letter reads. “This is essential to creating opportunities for innovation and prosperity across the country.”
The letter states that open weight will make it easier for new companies to enter the AI industry and increase competition that will strengthen the economy. According to the letter, Openweight gives customers greater control by allowing them to easily move between AI providers.
And while open weights pose “real and tangible risks,” anyone can change the AI model after release, leading to unregulated weights. Potentially malicious variantsThey also “expand defensive capabilities, increase transparency, and enable the discovery and remediation of vulnerabilities across many teams,” the letter signatories wrote.
The companies call the open weight AI model “one of the most important paths to AI safety and security.” They noted that closed-source models can be hacked or exploited in ways that are difficult to detect, and argued that concentrating the U.S. advanced AI market in the hands of a few large closed-source model providers “further exacerbates the risks.”
The letter’s signatories include some of the most influential people in the technology industry, including startups like Andreessen Horowitz. Technology giants like Cisco, Google, and Microsoft. And OpenAI.
Focus on security benefits
The Open Secure AI Alliance further emphasized cybersecurity in its launch announcement.
“Just as open source created a shared foundation for software, the United States and its partners now have a choice to make in AI security: Do we place the defenses that protect our infrastructure within a small number of opaque systems, or do we build on open models, harnesses, and tools that defenders can research, adapt, and deploy?”
The alliance noted that China’s open source model, GLM 5.2, has proven essential for cyber defense when using two OpenAI models. Hugging Face autonomously hacked an AI library And the library operator had to analyze the situation. The closed-source AI model refused to analyze the threat data input by Hugging Face, assuming it was attempting to launch its own attack. Only GLM 5.2, which Hugging Face was able to run on its own servers, answered the company’s questions and provided the defensive recommendations it needed.
Hug Face was one of 37 companies that helped create this partnership. Others include Adobe, CrowdStrike, NVIDIA, Palo Alto Networks, and Siemens.
As in the open letter, the Alliance acknowledged that there were real risks. Open source AI modelsincluding a lack of guardrails around cyber attack support. But “the right response is not to deny defenders access to capable open systems,” the alliance said. “It is important to combine openness with strong safeguards, clear rules against malicious exploitation, rigorous assessment, and rapid remediation.”
Cyber defenders need both open-source and closed-source AI models to protect critical infrastructure and other critical computer systems, the coalition said.
“Overall restrictions on open frontier AI systems risk weakening defensive capabilities and concentrating power, dependence, and vulnerability in a small number of closed providers,” the members wrote.
