Back to Blogs
Trust but Verify
The new AI executive order asks frontier developers to open their models to government review. We say, “Let’s give them the tools to self-evaluate prior to the government review.”
Seekr founder and CEO Pat Condo joined Newsmax on June 3 to discuss the new executive order. This piece expands on that conversation.
This week the President signed an executive order asking the companies building the most advanced AI to do something simple: let the government look at a model before it goes out the door. I talked through what it means on Newsmax this week – here’s the fuller version of what this EO means:
The review is voluntary. The window is short – up to 30 days. And it’s a request, not a requirement: a company doesn’t need the government’s approval to release a model. Most of the reaction has split along a familiar line. One camp calls it overreach, while the other calls it overdue, but both framings miss the more useful point sitting inside the order.
The order rests on a sound instinct: a powerful AI system should be examined before it’s turned loose. The next logical question is then who does the examining, and when? And the honest answer is that it should start long before a model ever reaches a government desk. It should start with the company that built it.
Think about how we used to handle software bound for sensitive environments. Before it went in, it was evaluated. And for the most sensitive systems, that meant examining the code itself to confirm nothing hostile had been planted. We didn’t just take the builder’s word for it.
A modern AI model is harder. Its behavior doesn’t live in lines of code you can read. It lives in billions of parameters you can’t. You won’t fix that by squinting harder at the weights. You fix it by building the system around the model so you can see what it’s doing, explain why, and – the part most tools skip – change it when it’s wrong. Watching isn’t enough. The system has to let you intervene: constrain the behavior, correct it, and steer it back. If we want to look before we deploy, the system has to be built to be looked into and acted on. Explainable and correctable by design – not opaque by default.
AI deserves the same discipline, and the builder should be the first one looking. Before a model goes anywhere, the company should run it through an agreed-upon set of evaluation tests by the government committee, then provide continuous updates to the model and agent builders so they can conduct a first pass self-evaluation: stress it, probe it, surface the red flags, and identify the places it breaks the rules it was given. Then it should be able to explain every one of those findings. Not, “The model seems fine,” but, “Here is what we tested. Here is what we found. Here is what we fixed.”
You can’t evaluate what you can’t inspect. You can’t defend a decision you can’t explain. You can’t correct a failure you can’t see. And an answer you can’t trace is an answer you can’t trust.
That is why explainability isn’t a feature you bolt on at the end. It’s a codified process using content, data, and model attribution, and attestability, that makes the testing possible in the first place. Our adversaries are building models shaped to serve their governments – tuned to censor, to slant, to follow the state’s line. The American answer isn’t to match their secrecy. It’s to build systems we can open up – models we can audit, interrogate, and hold to account, and hold others to the same standards. This accelerates innovation and adoption.
No one uses a system they can’t trust. Evaluation tools and systems arm and assist the human in the loop. A well-tested, explainable and transparent system codifies our laws and ethics, providing the guardrails and guidelines required to provide operational deployment with minimal risk.
At Seekr, we build AI you can trust.
Stop paying for confidently wrong answers
Wrong answers cost the same as right ones. Remove the hallucination tax with AI you can explain from the start.
Score your AI spend
Accelerate your path to AI impact
Book a consultation with an AI expert. We’re here to help you speed up your time to AI ROI.
Request a demo
