Tech companies must be primarily responsible for AI safety, Mike Johnson say
By [Author Name], Industrial Compliance Analyst at Gammatek ISPL
Last updated: September 2026 | 9 min read
Why This Matters Right Now
This week, the man second in line to the U.S. presidency said something that should matter to anyone running a business that touches AI: Congress isn't going to be the one setting the rules. House Speaker Mike Johnson told CNN's Jake Tapper on Sunday that he supports meeting with tech leaders on AI, but that developers — not lawmakers — are responsible for making sure their products are safe. If you're waiting for federal regulation to tell you what "responsible AI use" looks like for your industry, this week's comments are a signal that wait could be a long one. That has real consequences for anyone in a regulated sector who assumed government rules were coming to define the bar for them.
What Johnson Actually Said
Speaking on CNN's "State of the Union," Johnson said he wants a "steady hand at the wheel" on AI policy, but stopped short of committing Congress to writing new AI safety rules. He argued that the country needs to handle AI "like we have others in the past" — regulating carefully enough to manage risk without, in his words, smothering American innovation. Johnson also pointed to competitive pressure from China as a reason to avoid moving too aggressively on restrictions.
Johnson's position followed a call earlier in the week from Anthropic CEO Dario Amodei, who published an essay arguing that AI companies should deliberately slow the pace of capability improvements to leave room for safety work to catch up — a position that drew public support from Sam Altman and Elon Musk. Johnson, notably, did not endorse this "pacing" idea as something Congress should mandate; instead, he said the responsibility for implementing it sits with the companies themselves, since — in his assessment — the major AI labs don't even agree with each other on what safety guardrails should look like.
The Other Side of the Debate
Johnson's comments didn't happen in a vacuum — they landed in the middle of a wider, fast-moving argument in Washington. House Democratic Leader Hakeem Jeffries has called for Congress to begin regulatory talks immediately, arguing lawmakers should act with urgency rather than deferring to industry self-governance. Former President Barack Obama, speaking at a private fundraiser this week, warned that AI could become genuinely dangerous without an urgent, clearly-defined management plan — a notably direct statement from a former president typically cautious about weighing in on live legislative fights.
Critics of Johnson's position argue that voluntary industry self-regulation has a mixed track record across other industries, and that leaving safety standards to companies competing against each other creates pressure to cut corners rather than slow down. Supporters of Johnson's approach counter that heavy-handed, fast-moving legislation risks locking in rules based on today's technology that could be obsolete or counterproductive within a year, given how quickly AI capabilities are changing — and that U.S. competitiveness against China's more centrally-directed AI industry is a legitimate strategic concern.
Both sides agree on one thing, notably: that AI's risks are rising. Where they disagree is on who should be setting the pace of the response, and how fast that response needs to happen.
Why "Self-Regulation" Is Doing a Lot of Work in This Debate
The phrase "tech companies must be primarily responsible" sounds straightforward, but it's carrying more ambiguity than it first appears to. Johnson himself acknowledged that AI companies "have very different ideas on what the guardrails should be," and that there's no consensus among them on safety standards — which raises an obvious question: responsible according to what standard, exactly, if the companies themselves can't agree?
This is where the debate over AI safety starts to resemble a pattern regulated industries have seen before. Financial services, pharmaceuticals, and industrial manufacturing all went through periods where the debate was framed as "should government regulate this, or should the industry police itself" — and in nearly every case, the eventual answer wasn't purely one or the other. It became a mix: baseline government standards, paired with industry-specific compliance frameworks that translate broad principles into auditable, specific practice.
An Implementation Consideration for Regulated Industries
For manufacturers, chemical plants, and pharma operations already operating under compliance frameworks like IEC 62443 or FDA/EPA regulatory standards, this debate isn't abstract — it's a preview of a decision your organization may face directly. If AI safety rules end up structured the way most regulated-industry compliance eventually does — baseline principles from government, translated into specific practice by the industry itself — the plants that already have strong internal compliance and audit infrastructure will adapt fastest, because they already have the documentation habits and internal ownership structures this kind of self-regulation requires.
Plants without that infrastructure will likely find themselves scrambling to build audit trails and accountability structures reactively, once new expectations (whether from Congress, sector regulators, or insurance/liability pressure) arrive — a pattern that's played out before with cybersecurity compliance and environmental reporting requirements.
What to Actually Watch Next
A few concrete developments worth tracking, rather than speculating on:
Whether Johnson's informal "talks with tech leaders" produce anything more binding than a public statement
Whether Jeffries and Congressional Democrats can force floor action given they don't control the chamber's agenda
Whether any of the major AI labs actually follow through on Amodei's "pacing" proposal, or whether it remains a rhetorical commitment
Whether existing regulators (FTC, sector-specific agencies) begin asserting authority over AI safety under current law, rather than waiting for new legislation — a path that requires no new congressional action at all
Where This Leaves Regulated Businesses Today
Whatever Washington eventually decides, the practical reality for manufacturers, pharma operations, and chemical plants integrating AI into their processes is the same either way: internal compliance and audit infrastructure isn't something to wait on legislation for. Building the systems to document, monitor, and prove responsible AI use now is the kind of preparation that pays off regardless of which direction federal policy eventually goes — and it's exactly the gap between "having a policy" and "being able to prove compliance" that trips up organizations when new requirements do arrive.
Comments