The thing that gets me is not the lobbying itself, it’s that they’re comfortable saying two completely different things depending on who’s listening. Sam Altman in an interview sounds like open source is the future. Sam Altman in a backroom with regulators sounds like open source models are a national security problem.
That gap matters because it tells you what they actually believe. If you genuinely thought open source was dangerous, you’d say so publicly and defend it.
source: https://www.nytimes.com/2026/07/25/technology/open-source-silicon-valley-china.html
The rhetoric gap is the less interesting part honestly. What matters is whether the compute thresholds they’re pushing for happen to sit just above what they themselves run and just below what a serious open source competitor would need.
Has anyone actually lined up the specific numbers in these proposals against current frontier training runs? If the threshold moves every time someone gets close to it, that tells you everything.
The concern here is less about whether open source AI is good or bad, and more about consistency and transparency. Technology companies often have to balance public messaging, business interests, and regulatory discussions, which can create differences in how they frame the same issue.
Open source models can bring huge benefits through innovation, research, and wider access, but there are also legitimate concerns around misuse, security, and accountability. The important part is that companies should clearly explain their position and the reasoning behind any policy recommendations rather than appearing to change views depending on the audience.
A more open discussion between industry, researchers, policymakers, and the public would help build trust around the future of AI.
Ask ClosedAI to publish the actual model weights they cite as “misuse risk” versus the ones they don’t, and see how fast that clarity vanishes.
