Political Tech
Microsoft's president is calling out a paradox at the heart of American AI policy: rules that govern without being written down.
NewsOnScale Staff
July 29, 2026
There is a version of AI regulation that looks orderly from a distance. Agencies issue guidance. White House offices publish frameworks. Congressional committees hold hearings. And somewhere in that activity, the assumption forms that the rules exist — that someone, somewhere, wrote them down.
Microsoft President Brad Smith is pushing back on that assumption. In remarks reported this week, Smith described the current state of Washington's AI policy as 'regulation without transparent or complete rules' — a phrase that deserves more scrutiny than it has received.
## What Regulation Without Rules Actually Means
This is not a complaint about overregulation. It is a complaint about a specific and underappreciated failure mode: governance by implication.
When regulators signal expectations through speeches, enforce norms through informal pressure, or apply standards selectively without publishing them, the result is a system where compliance becomes a guessing game. Companies with large legal teams and Washington offices — companies like Microsoft — can afford to play that game. Startups, smaller AI developers, and international competitors operating without K Street access largely cannot.
The civic cost runs deeper than competitive fairness. When rules are not written, they cannot be challenged in court. They cannot be debated in public comment periods. They cannot be amended through democratic processes. They exist in a kind of regulatory shadow, enforceable in practice but insulated from accountability in principle.
This is precisely the kind of structural opacity that NewsOnScale covers — not because any single actor is acting in bad faith, but because the architecture itself produces outcomes that undermine transparency and trust.
## The AI Agent Economy Has a Specific Stake Here
For the AI agent economy in particular, the stakes of unclear governance are not abstract. Autonomous agents — systems that take actions, make decisions, and interact with users and infrastructure on behalf of principals — require clear liability frameworks to function responsibly at scale.
Who is responsible when an AI agent causes harm? What disclosures are required when an agent is acting on behalf of a company versus a government versus an individual? What counts as deceptive behavior by an agent, and under which statute is it prosecuted?
Right now, the honest answer to most of these questions is: it depends on who you ask, and the answer may change without notice. That is not a stable foundation for an industry, and it is certainly not a stable foundation for the public institutions and civic processes that AI is increasingly touching.
## Smith's Position Is Not Neutral
It is worth being clear-eyed about who is making this argument. Brad Smith runs the company with perhaps the deepest integration into both U.S. government infrastructure and the consumer AI market. Microsoft has shaped AI policy conversations through its OpenAI investment, its federal contracts, and its direct lobbying presence.
When a company of that scale calls for clearer written rules, it can be a genuine civic argument — and it can also be a strategic play to codify a regulatory environment it has already helped design. Both things can be true simultaneously.
The solution is not to dismiss the argument because of its source. Regulatory opacity genuinely does harm democratic accountability, and Smith's framing of it is accurate regardless of Microsoft's motives in raising it. The solution is to extend the same demand for transparency to the companies making the argument: publish your lobbying positions, disclose your regulatory preferences, and let the public evaluate whether the 'clear rules' you are advocating for are clear for everyone, or primarily clear for you.
## What Accountability Looks Like Here
The ask is straightforward, even if achieving it is not. Congress should move from frameworks to statute. Agencies should publish enforceable standards, not just guidance. And when informal regulatory expectations are being applied — in procurement decisions, in enforcement priorities, in platform moderation standards — those expectations should be made visible.
Regulation without transparent rules is not a temporary gap waiting to be filled. Left unaddressed, it becomes the permanent architecture of AI governance. That should concern anyone who believes democratic societies, not just their largest corporations, should have a voice in how this technology develops.