Can ‘super intelligence’ and a non-binding safety pact solve AI’s image problem?

There's a peculiar rhythm to how governments try to manage transformative technology: lead with the awe, follow with the asterisks. The phrase "super intelligence" is doing heavy lifting here—it's a term that evokes sci-fi wonder and, crucially, shifts the conversation away from the messy, immediate harms of today's AI (bias, disinformation, job displacement) toward a speculative, almost theological future. A non-binding safety pact, meanwhile, is the policy equivalent of a gentle hand on the shoulder. It signals concern without the inconvenience of enforcement. This isn't a new move. We've seen voluntary frameworks from major AI labs that crumble the moment market pressure meets safety pledges. The administration's approach appears to borrow that same playbook but with a political twist: rebrand the technology as a patriotic, world-beating ambition while simultaneously appearing to "do something" about safety. The problem is that the public has become remarkably literate in spotting the difference between a commitment and a commitment device. The deeper question the episode seems to circle is whether trust can be manufactured through messaging at all. Trust in AI is currently a function of lived experience—people have seen chatbots hallucinate, seen deepfakes, seen automated systems fail. No summit communiqué or catchy slogan rewrites that record. If the administration's goal is genuinely to improve AI's image, the fix isn't a new word for intelligence; it's a demonstrable mechanism for accountability that outlives any single administration. Source: [TechCrunch](https://techcrunch.com/2026/10/04/can-super-intelligence-and-a-non-binding-safety-pact-solve-ais-image-problem/)