Microsoft signed a binding contract with America’s second-largest teachers’ union. A Louisiana senator is bringing his own kill-switch bill to the floor this week. And the president’s answer to both: “The only guardrails AI needs is a STRONG AND SMART (High IQ!) PRESIDENT.”
Two distinct AI guardrail efforts advanced, and the president dismissed both. Microsoft signed a legally enforceable AI safety standard with the American Federation of Teachers on 9 September, restricting how its AI behaves inside US classrooms. Days later, President Trump posted that AI safety concerns amount to a “conspiracy,” writing on Truth Social: “The only control or ‘guardrails’ that AI needs is a STRONG AND SMART (High IQ!) PRESIDENT, and the U.S.A. has that, in spades!” Both developments are inside TF’s coverage in AI Slowdown Debate Reaches Wall Street, Congress, Westminster — a moment where private-sector guardrails and congressional pressure are moving in one direction. In contrast, the White House is moving in another.
What’s Happening & Why It Matters
Microsoft: Audit Us
Here’s what separates Microsoft’s agreement from a typical corporate pledge. The standard, negotiated with AFT President Randi Weingarten and UFT President Michael Mulgrew, prohibits AI companions or features “designed to foster emotional attachment or dependency,” and bars anything engineered to keep students engaged longer than a learning task requires. Microsoft has committed not to use student or educator data to train its AI systems, apart from a narrow exception covering data that could help protect students. Crucially, the standards are binding and require third-party audits to verify compliance — not a voluntary commitment Microsoft can walk back.
Starting 1 November, every US school district and independent school with an existing Microsoft agreement can add the protections without renegotiating their contracts. Microsoft Vice Chair Brad Smith stressed that teacher judgment must stay central, not replaced by AI. Josh Golin, executive director of online safety nonprofit Fairplay, offered a pointed qualifier: the provisions are meaningful, but effective only “if all AI companies, especially the tech giants, sign on.” OpenAI and Anthropic have both said they’re discussing comparable pacts with the AFT — but Google, the dominant provider of education technology in American schools, has not said whether it plans to offer similar protections.
NYC and LA Press Pause
The agreement is part of a pattern TF has already covered. As reported in NYC Bans AI Use in Schools Until High School, New York City paused student AI use for a full year at the start of this school year. Los Angeles, the country’s second-largest district, made the same call. Both moves are districts deciding the honest answer to “how should AI work in classrooms” is “not yet, not like this.” Microsoft’s contract doesn’t reverse that caution — it answers a narrower question for districts that keep using Microsoft’s tools regardless: what specific behaviours are off-limits while AI is present.
A separate Austrian study cited in this week’s coverage found 60% of surveyed 11- to 17-year-olds had already sought advice from AI chatbots about stress, conflict, or heartbreak, with 30% discussing worries or feelings. That’s the exact “emotional attachment or dependency” risk Microsoft’s contract targets — evidence the behaviour already exists at scale, not a hypothetical the agreement is getting ahead of.
Kill Switch Bill Reaches the Senate Floor
Congress isn’t waiting for Microsoft’s private-sector template either. Senator John Kennedy (R-La.) confirmed he plans to bring a kill-switch measure to the floor Wednesday, requiring AI developers to maintain the technical capacity to shut down their own systems if needed — a federal version of the UK proposal TF covered being rejected in UK Govt: No Kill Switch ‘Cause We Can’t Turn AI Off. “AI is becoming an issue, particularly in the last few days,” Kennedy said. Representative Don Beyer (D-Va.), who leads the Congressional Artificial Intelligence Caucus, said he believes sufficient support exists for multiple bills, including a bipartisan measure establishing AI model guardrails alongside Kennedy’s kill-switch provision.
Polling gives that bipartisan structure real weight. A University of Maryland Program for Public Consultation survey found 85% of Democrats and 79% of Republicans want a new federal agency dedicated to AI oversight. In this rare policy area, voters in both parties agree, even as their elected leadership splits on what to do about it.
“Conspiracy” Splits White House, Republicans
Trump’s response went beyond social media posts. He made an impromptu call to Nvidia CEO Jensen Huang, who was onstage at a technology conference at the time, putting their conversation on speakerphone in front of the audience — an unusual, theatrical show of alignment between the White House and the AI industry’s most visible chip supplier. House Science Committee Republicans echoed Trump’s view, posting about their “ongoing investigation into potential foreign influence behind organised opposition to U.S. AI data centre development” — recasting domestic safety concerns as a possible foreign influence operation rather than a genuine policy debate.
Not every Republican is following that line. House Speaker Mike Johnson posted a more measured position Sunday night: “We all have a sense of urgency to create guardrails around the technology. The key is designing the guardrails, in a way that prevents any harm from AI — while also preserving American innovation and our national security.” Johnson reiterated that position to reporters Monday. However, he also called Congress’s typical instinct to “cover things up with red tape” a problem in itself. He suggested Trump summon AI leaders for a meeting rather than pursue legislation. Senator Mark Warner (D-Va.) rejected that stance, arguing voluntary industry pledges “do not meet this moment” and urging Congress to “act this week.”
Scepticism Runs Deep
The most honest assessment of Congress’s capacity came from Senator Rick Scott (R-Fla.), who questioned whether meaningful AI legislation is realistic given Congress’s track record: “Have we gotten election security done? Did we balance a budget? Oh, did we pass a budget? See, what have we gotten done? So you think we’re going to get AI done?” Senator Mark Warner offered a blunter diagnosis of the pattern: “We never do anything until we have a holy heck moment.” A 2024 bipartisan AI working group formed under then-Senate Majority Leader Chuck Schumer recommended $32 billion in AI development and safeguard spending — with little meaningful follow-through since.
The history factors into the momentum. The House is scheduled to adjourn at the end of the week. It won’t return until after the November elections, leaving a narrow window for any of this to translate into a vote before the midterm calendar kicks in.
TF Summary: What’s Next
Kennedy’s kill-switch measure faces a Senate floor vote Wednesday and needs full chamber support to advance. Microsoft’s AI Safety and Privacy Standard is available to every US school district on 1 November. OpenAI and Anthropic’s own discussions with the AFT over comparable school-specific pacts are ongoing, with no confirmed agreement date. The House adjourns for the year at week’s end and will not return until after November’s elections.
MY FORECAST: Expect Kennedy’s kill-switch bill to fail this specific floor vote, given Congress’s own documented pattern of narrow AI measures collapsing even with declared bipartisan support — Senator Scott’s scepticism is an accurate read of the chamber’s recent legislative record, not just partisan posturing. Microsoft’s school contract will prove the more durable outcome this week, because it doesn’t require Congress to act at all — a binding, audited private agreement sidesteps the exact gridlock Scott and Warner both described. Watch whether Google faces mounting pressure to match Microsoft’s terms once more districts start asking why the dominant classroom technology provider hasn’t offered comparable protections; Fairplay’s own critique already identifies that gap as the standard’s most obvious weakness.
Related Stories
- AI Slowdown Debate Reaches Wall Street, Congress, Westminster
- NYC Bans AI Use in Schools Until High School
- UK Govt: No Kill Switch ‘Cause We Can’t Turn AI Off