It’s not about Anthropic vs. OpenAI anymore

The U.S. government is poised to significantly influence the release of AI models, creating a new dynamic in the tech industry. Recently, Anthropic’s Fable and Mythos models were halted, and now it appears OpenAI's upcoming GPT 5.6 may face a similar fate. Reports indicate that its rollout will be limited, with the government approving each release on a customer-by-customer basis until a broader approval is granted. If this limited preview period is brief, as suggested by Altman, it may not pose a substantial problem. However, with Mythos already languishing in preview for months, the prospect of further delays raises concerns over the economic viability of new AI systems. A prolonged review process could hinder the financial performance of AI labs that are under pressure to enhance their profits. A slowdown in model development could also negatively impact the ongoing expansion of data centers, potentially putting the entire sector at risk. Both OpenAI and Anthropic find themselves confronting similar obstacles, with a looming disaster if they fail to navigate this new landscape. Discussions within the tech community often revolve around the competing interests of these companies, with accusations of regulatory manipulation and political favoritism. Yet, the issue at hand transcends individual company rivalries. Implementing an inconsistent government approval process for every new AI model presents clear challenges. While it is reasonable for the government to review models before their release, as is customary for various consumer products, questions arise regarding the safety assurances necessary to satisfy regulatory standards. The U.S. government currently lacks the expertise to conduct the level of testing required, and the specific risks that regulators aim to mitigate remain unclear. Beyond the complications of the regulatory process lies a pressing need to address legitimate concerns around AI technologies. Even skeptics of Mythos cannot deny the transformative impact AI tools are having on cybersecurity and other fields. Therefore, limiting model releases cannot be the singular solution; it risks restricting public access to innovations. To effectively address these challenges, collaboration is essential. This involves trusting independent organizations to guide the regulatory process, even if their goals don’t fully align with corporate interests. It requires a unified approach to embrace the least problematic regulatory options rather than resisting all regulations vigorously. Most importantly, it calls for a collective commitment to advancing the AI industry, recognizing that safety and regulation can be approached as collaborative opportunities rather than competitive advantages. As AI models continue to evolve and their capabilities yield significant political implications, the industry must engage in collective action. The coming weeks will reveal whether the tech sector can rise to this challenge.