The most ambitious legislative priority in Silicon Valley's Washington playbook - a single federal law on artificial intelligence that would wipe out the growing patchwork of state regulations - is running out of road, and it may be taking a popular child safety bill down with it. With the midterm elections approaching and the legislative calendar already overloaded, lobbyists for major technology companies are watching their best remaining opportunity for federal AI preemption collapse under the weight of procedural confusion, fractured alliances, and a last-minute political marriage of convenience that almost no one in Washington appears to have agreed to.
What Preemption Means - and Why Big Tech Wants It So Badly
For technology companies operating at national scale, a state-by-state regulatory environment is an operational and legal headache. Over the past several years, states including California, Colorado, Texas, and Illinois have each moved forward with their own rules governing automated decision systems, algorithmic accountability, and data handling - producing a compliance landscape that varies widely in scope and strictness. Federal preemption would resolve that problem with a single stroke: one law, one standard, applied uniformly, with state laws either subordinated or invalidated.
That goal has defined the industry's posture in Washington for months. But the path to preemption requires passing something through Congress and getting it signed by the president, neither of which has proven easy. Republican populists in the House have been skeptical of legislation that appears to shield large technology companies from accountability. Democrats have resisted any framework that trades meaningful state-level protections for a weaker federal floor. The result has been a deadlock that lobbyists have struggled to break.
President Donald Trump has called for passage of an AI preemption bill, which means Republican leadership cannot simply ignore the issue. The White House's current approach has been shaped in part by Mike Davis, a Trump-allied lawyer and founder of the Article III Project, who helped kill a separate AI moratorium in the Senate last year. Davis has argued publicly that any preemption legislation must address what he calls the "Four Cs" - children, conservatives, creators, and communities. Without satisfying all four, he has said, the bill will not move. "There is no chance in hell AI preemption will pass if it does not address the Four Cs," Davis told The Verge. "I will make damn sure of that. Again."
A Deal No One Agreed To
Earlier this week, reports emerged that the White House had signaled its support for bundling AI preemption with the Kids Online Safety Act, or KOSA, a children's online safety bill championed by Sen. Marsha Blackburn (R-TN). KOSA has a genuine legislative history: the Senate passed it 91-3 in 2024, with bipartisan support led by Blackburn and Sen. Richard Blumenthal (D-CT). On its face, attaching a popular child safety bill to an unpopular industry priority looks like an obvious political calculation - use KOSA's goodwill to carry preemption across the finish line.
The problem is that the White House apparently executed this maneuver without informing the House Republicans, who had passed their own version of KOSA, or the Senate Democrats who co-sponsored Blackburn's version. Both groups found out about the arrangement at roughly the same time the public did. The resulting confusion has been significant. House and Senate versions of KOSA differ substantially: the Senate version requires technology companies, including AI developers, to assume a formal "duty of care" toward young users - a proactive obligation to prevent foreseeable harms. The House version, advanced by Majority Leader Steve Scalise (R-LA), diluted that provision considerably last November, drawing sharp criticism from child safety advocates. Blackburn's preference for the Senate's stricter language is not a minor procedural disagreement; it is the core of the bill's protective value, and Scalise's camp has no obvious incentive to accept it.
"No one knows really who's driving this thing," a Republican lobbyist for a midsize tech company told The Verge. "Everyone is deeply, deeply, deeply skeptical of its movement, because everyone is on such different pages. I think the House is not going to move anything that Blackburn wants."
The Calendar Problem
Even if the political divisions could be smoothed over, the legislative schedule presents a near-insurmountable obstacle. It is now mid-June. Congress breaks for a five-week recess in approximately six weeks. After that, the calendar is consumed by general election dynamics, at which point meaningful bipartisan cooperation on contested legislation becomes functionally impossible.
The weeks remaining are already spoken for. Congress must address the renewal of FISA surveillance authorities, an immigration enforcement package, defense spending increases tied to the administration's Iran policy, a crypto market structure bill, affordability measures, the SAVE America election legislation, and routine budget obligations including Medicaid funding. Fitting a comprehensive AI law - one that also reconciles competing versions of a child safety bill - into that window strains credulity. "It is mid June. You have a month and a half before people leave for recess. And then it's election season," one AI policy advocate told The Verge. "There's just no way."
The post-election calculus makes things worse, not better. If Democrats retake even one chamber in November, their incentive to cooperate on legislation that delivers Big Tech's top regulatory priority is essentially zero. A new Congress would mean new legislative priorities and, almost certainly, a far more adversarial stance on AI preemption. "After the election, what incentive do the Democrats have to support anything?" the Republican lobbyist asked. "I'm deeply skeptical."
An Industry Forced Into a Corner
The deeper strategic problem for the technology industry is that the KOSA attachment has introduced a genuine values conflict. Federal AI preemption, as the industry wants it, would limit the liability exposure of AI companies under a uniform federal standard. A robust "duty of care" provision of the kind in the Senate's KOSA would create exactly the kind of preemptive legal obligation that the industry has resisted. The two goals are not obviously compatible. Accepting KOSA in its stronger Senate form in order to secure preemption would mean accepting a meaningful legal duty toward young users - a standard that could be applied broadly as AI becomes more embedded in consumer products.
Austin Carson, founder of SeedAI and formerly Nvidia's head of government relations, was blunt about the prospects. "I can't imagine a scenario where this bill would move," he told The Verge. "I cannot imagine it." That skepticism is widely shared. The legislation is simultaneously trying to satisfy a president who wants AI preemption, a Trump-allied gatekeeper who wants the Four Cs addressed, House Republicans who want their version of KOSA, Senate Republicans who want theirs, Senate Democrats who co-sponsored the original bill but didn't agree to the new arrangement, and child safety advocates who have spent years fighting for the duty-of-care standard the House tried to remove.
The irony is considerable. Big Tech spent months in pursuit of preemption precisely because a fragmented regulatory environment felt unmanageable. The attempt to resolve that fragmentation has produced, at least for now, a different kind of disorder - one that is political rather than legal, and considerably harder to litigate away.
