Sanders Goes for the Kill Switch

While the Pentagon-Anthropic lawsuit grinds through closed-door negotiations, Sen. Bernie Sanders (I-VT) dropped a legislative proposal that will reshape the AI debate for years.

Sanders announced legislation in 2026 to impose a full moratorium on new AI data center construction until "strong national safeguards are in place," according to Reason.

"Super intelligent AI could become smarter than human beings, could become independent of human control, and could pose an existential threat to the entire human race," Sanders said in a video posted to social media.

The same senator who wants government to run your healthcare now wants government to pump the brakes on the single most transformative technology since electricity.

The problem? Freezing American AI development doesn't freeze China's. A moratorium is a unilateral disarmament disguised as safety policy. Reason correctly points out that equivalent logic in 1889 would have stopped the electrification of America over a dead Western Union lineman.

Anthropic's CEO Dario Amodei takes a different lane. In a 2026 essay cited by Reason, Amodei wrote: "We are so close to these models reaching the level of human intelligence, and yet there doesn't seem to be a wider recognition in society of what's about to happen." He's calling for "sensible AI regulation," not a freeze.

One approach tries to steer the car. The other throws a brick through the windshield.

Courts Just Changed the Rules for Expert Witnesses

On Monday, Judge Thomas O. Farrish of the U.S. District Court in Connecticut ruled in Conservation Law Foundation, Inc. v. Shell Oil Co. that AI prompts used by an expert witness are subject to compelled discovery under Rule 26(b), according to the Volokh Conspiracy.

The case: Shell Oil's defense team demanded that plaintiff expert Dr. Naomi Oreskes hand over the AI prompts she used to analyze Shell's document production. Conservation Law Foundation argued the prompts were off-limits — claiming they were "notes" protected under a prior discovery agreement, or simply not within discovery scope at all.

Judge Farrish rejected every argument.

His reasoning is blunt: an expert's methodology is fair game for discovery. If you used AI to cull through thousands of documents and identify which ones matter, the prompts you fed that AI are part of your methodology. Full stop.

CLF also tried to reframe the prompts as mere "search terms" — claiming it had already produced those. The judge wasn't impressed. He indicated that whether the prompts truly don't exist, or whether CLF is playing word games, would need to be resolved.

Any expert witness using AI tools in federal litigation now has to assume their prompts are discoverable. That changes how law firms build cases, how experts conduct analysis, and how AI outputs get scrutinized in court.

Mainstream legal media has barely touched this.

The Compliance Gap Is Real and Getting Dangerous

Meanwhile, corporate America is running AI tools faster than any governance structure can track them.

The Workplace Privacy Report's top compliance issues for 2026 lead with a stark warning: AI governance will be judged "less by aspirational principles and more by documented processes, controls, and accountability." Organizations are expected to maintain enterprise AI inventories — including so-called "shadow AI" embedded in existing software — and classify systems by risk.

Most companies are NOT doing this. The report specifically flags AI-driven employee monitoring tools — dashcams, performance management software, wearables — as a litigation minefield. Regulators and plaintiffs are zeroing in on whether AI monitoring tools collect more data than necessary, and whether the outputs are explainable when they influence hiring or firing decisions.

IBM's analysis, co-authored by Alexandra Jonker and Alice Gomstyn, puts the privacy stakes plainly. Jennifer King, a fellow at Stanford University's Institute for Human-Centered Artificial Intelligence, told the institute: "We've seen companies shift to this ubiquitous data collection that trains AI systems, which can have major impact across society, especially our civil rights."

Privacy matters across the political spectrum. Both government and corporations have histories of abusing what they collect.

What Mainstream Coverage Is Getting Wrong

Most outlets are treating the AI governance fight as a culture war proxy — either cheerleading for Silicon Valley against government overreach, or painting Big Tech as a surveillance monster that needs to be caged.

Both framings miss the actual stakes.

Anthropic was right to push back on the Pentagon using its AI for autonomous weapons and domestic surveillance, according to Reason's reporting. Defense Undersecretary Emil Michael's complaint — that Anthropic was "inserting policy preferences in the middle of an operation" — sounds like a bureaucrat mad that a vendor has a conscience.

But Sanders' moratorium is the wrong answer. Stopping American AI development hands the advantage to Beijing.

And Judge Farrish's discovery ruling is a preview of where this is all heading: into courtrooms, where the absence of clear rules will be settled one lawsuit at a time, at enormous cost to everyone involved.

The Accountability Vacuum

If you work at a company using AI tools — and you almost certainly do — your employer may not know what data those tools are collecting about you. The Workplace Privacy Report says most organizations haven't audited their monitoring tools for AI functionality. That's your performance review data, your communications, potentially your location.

If you're involved in litigation and your side's expert used AI to build the case, that analysis is now an open book for the other side.

If Congress can't get past the binary of "ban it" vs. "let it rip," the courts and the regulators will fill the void — messily, expensively, and too slowly to matter.

The technology is NOT waiting for Washington to figure it out.