logan koepke

Brown University researchers launch AI legislation tracking and analysis portal As a tool for researchers, lawmakers, journalists and the public, the CNTR AISLE Portal provides analysis of state- and federal-level AI bills pending across the U.S.

Great resource for anyone engaged or interested in US AI policy and governance from @geomblog.bsky.social and his team www.brown.edu/news/2026-03...

my hope is that the events of the past few days will be catalytic in driving advocacy and results in legislative changes that ultimately protects real people from the material harms of these systems.

this should unite AI advocacy communities that have often been at odds.

once you are in the "all lawful purposes" game, it is effectively carte blanche.

the national security law stack all but ensures that the powers that be will be able to use systems in manners they see fit that comport with their long-standing view: rampant surveillance.

ultimately, the ongoing contractual saga between Anthropic, OpenAI, and the DOD highlights how important it is for congress to establish clear rules prohibiting mass surveillance and lethal AI use.

the entire game is "all lawful purposes." and the laws enable mass surveillance.

How Talks Between Anthropic and the Defense Dept. Fell Apart

NYT’s reporting on the DOD-Anthropic dispute sheds more light on how the Pentagon planned to use Claude for mass surveillance of Americans: by collecting and analyzing commercial bulk data.

So what is this commercial data? And how might Claude be used to collect and analyze it? 🧵 1/

OpenAI’s Pentagon red lines are a mirage OpenAI claims its DoW deal prevents its models being used for mass domestic surveillance. That appears to be misleading at best

this piece does a good job of recapping this weekend's events www.transformernews.ai/p/openai-pen...

Sen. Ossoff Statement on Iran Atlanta, Ga. - U.S. Senator Jon Ossoff today released the following statement. "Eight months ago, Donald Trump lied to the country when he claimed U.S. air strikes had 'obliterated' Iran's nuclear program. "Now Trump says he's taken America to war for regime change in Iran. "Trump has launched this regime change war and put American forces at risk without presenting evidence of an imminent threat, without clear objectives, without having exhausted diplomacy, without a plan for the aftermath, and without the consent of Congress. "I oppose yet another regime change war-of-choice in the Middle East. Congress must convene immediately and pass a War Powers Resolution to assert our Constitutional authority over war and peace. "The President has presented no evidence that Iran or its nuclear program — which he falsely claimed to have 'obliterated' — pose an imminent threat to the United States, or that diplomacy was exhausted. Iran's support for terrorism and uranium enrichment have long destabilized the region. But sending American forces into harm's way should only ever be a last resort. "I pray for the safety of U.S. forces who have been ordered into battle and commend their extraordinary professionalism."

Ossoff’s statement is 1000x stronger than Schumer-Jeffries. This is a moderate Jewish Senator from a Trump state. The split isn’t moderate vs progressive, but leading vs poll-reacting

So, so instructive to compare media response to Afghanistan withdrawal -- months of hysterical garment-rending & catastrophizing, relentless criticism of the admin -- with the response to an unprovoked, unmotivated, illegal war of aggression, ie, "some critics say maybe Trump should have reasons."

As Friday’s deadline approached, Anthropic executives thought they were close to a compromise with the Pentagon and were just a few words apart on the issue of surveillance, people on both sides of the negotiation said. Complicating the matter was a social media post by President Trump. Mr. Trump had told Mr. Hegseth on Friday morning that he had prepared a post belittling Anthropic and ordering all government agencies to stop working with it within six months. Even as Mr. Trump published the post at 3:47 p.m., the two sides kept talking. Mr. Michael, who was on a call with Anthropic executives at the time, said the Pentagon wanted the company to allow for the collection and analysis of unclassified, commercial bulk data on Americans, such as geolocation and web browsing data, people briefed on the negotiations said. Anthropic told the Pentagon that it was willing to let its technology be used by the National Security Agency for classified material collected under the Foreign Intelligence Surveillance Act. But the company wanted a legally binding promise from the Pentagon not to use its technology on unclassified commercial data.

"the Pentagon wanted the company to allow for the collection and analysis of unclassified, commercial bulk data on Americans, such as geolocation and web browsing data, people briefed on the negotiations said."

www.nytimes.com/2026/03/01/t...

• perhaps DOD will claim their understanding of the contract also aligns with OpenAI and that things are keyed to laws and policies as they stand today. but, uh ... quite recent events, suggests DOD cares a lot about preserving their understanding of "all lawful purposes"

What if the government just changes the law or existing DoW policies? Our contract explicitly references the surveillance and autonomous weapons laws and policies as they exist today, so that even if those laws or policies change in the future, use of our systems must still remain aligned with the current standards reflected in the agreement.

OpenAI suggests that their contract dates use to laws and policies as they exist today. but, this seems tenuous given:

• the core premise of the contract is "all lawful purposes" (seems circular)
• the nature of 3000.09, which DOD can change at anytime

How the U.S. Military Buys Location Data from Ordinary Apps The U.S. military is buying the granular movement data of people around the world, harvested from innocuous-seeming apps, Motherboard has learned. The most popular app among a group Motherboard analyzed connected to this sort of data sale is a Muslim prayer and Quran app that has more than 98 million downloads worldwide. Others include a Muslim dating app, a popular Craigslist app, an app for following storms, and a “level” app that can be used to help, for example, install shelves in a bedroom.

Intelligence Analysts Use U.S. Smartphone Location Data Without Warrants, Memo Says The disclosure comes amid growing legislative scrutiny of how the government uses commercially available location records.

WASHINGTON — A military arm of the intelligence community buys commercially available databases containing location data from smartphone apps and searches it for Americans’ past movements without a warrant, according to an unclassified memo obtained by The New York Times. Defense Intelligence Agency analysts have searched for the movements of Americans within a commercial database in five investigations over the past two and a half years, agency officials disclosed in a memo they wrote for Senator Ron Wyden, Democrat of Oregon.

DOD Components acquire, access, and use information that is available to the American public and consumers worldwide to plan, inform, enable, execute, and support a wide range of missions lawfully and responsibly, including the Department's foreign intelligence and cybersecurity missions, security activities, and to protect DoD personnel and information from foreign adversary threats. These activities are conducted in accordance with all applicable laws, including the Fourth Amendment to the Constitution, the Foreign Intelligence Surveillance Act, the Privacy Act, and DoD's implementing policies. I am not aware of any requirement in U.S. law or judicial opinion, including the Supreme Court's decision in Carpenter v. United States , 138 S. Ct . 2206 ( 2018 ), that DoD obtain a court order in order to acquire, access, or use information, such as CAI, that is equally available for purchase to foreign adversaries , U.S. companies, and private persons as it is to the U.S. Government. DoD Components acquire and use in a manner that adheres to high standards of privacy and civil liberties protections, and that accords with DoD's national security missions

of course, the DOD has purchased bulk commercial data for surveillance and has argued nothing prohibits them from acquiring such data. quite unclear what technical methods are supposed to help prevent LLM analysis of bulk commercial data

www.vice.com/en/article/u...

www.nytimes.com/2021/01/22/u...

1. NSA engages in incidental domestic collection under FISA 702 and makes it available for queries, and DoJ writes an annual report to Congress listing all the times it does. OpenAI models usable under your contract for that, yes or no? 2. Getting and/or analyzing commercially available data at scale. OpenAI models usable under your contract for that, yes or no? 3. OLC writes a "President's Surveillance Program 2.0"-like memo claiming President has inherent CinC authority to authorize mass domestic warrantless wiretapping. OpenAI models usable under your contract for that, yes or no? and, 4., how can we verify that? -- 1. No, this contract does not apply to NSA. 2. The Pentagon has no legal authority to do this (that would be federal law enforcement agencies, not DoW) 3. Again, if this were to happen (and to my knowledge it hasn't) this could only be done by the FBI and they are not a party to this contract. 4. Read the authorizing statute for the Department of Defense? None of these activities are within their statutory authorities. And our contract is expressly limited to the Department of Defense.

We can't protect against a government agency buying commercially available data sets, but our contract incorporates a prohibition on mass domestic surveillance as a binding condition of use.

OpenAI claimed two things yesterday:

1. DOD "has no legal authority" to buy commercially available data (they have done this in the past and argue they do)
2. that "we can't protect against a government agency buying commercially available data sets."

a bit contradictory

the clause (which is a legitimating device, not a constraining device) in some sense presupposes that we have a functioning system.

in other words: if the provision is aimed to mean that the company's system can be used for any purpose that is lawful under applicable law, how does that apply when the President is prosecuting an illegal war?

the US is now engaged in an illegal war against Iran. Congress has provided no authorization or declaration.

given that: is the fight about DOD "all lawful purposes" language for AI companies really just downstream of questions about engaging in an illegal war?

Preventing Woke AI in the Federal Government By the authority vested in me as President by the Constitution and the laws of the United States of America, it is hereby ordered: Section 1.  Purpose.

also helped write the "Preventing Woke AI in the Federal Government" Executive Order. glad he recognizes this is bad. but his work has to have helped pave the way for this.

www.whitehouse.gov/presidential...

Exclusive | Government Agencies Raise Alarm About Use of Elon Musk’s Grok Chatbot Warnings about xAI’s safety and reliability preceded the Pentagon’s decision to approve Grok for use in classified settings.

i'm baffled as to what the next step is supposed to be.

if OpenAI or Google attempt to articulate a similar red line (potentially an if here), does the DOD attempt to do this all again?

they know they cannot seriously rely on xAI where "safety is a dead organization."
www.wsj.com/politics/nat...

We Will Not Be Divided Employees of Google and OpenAI stand together to refuse the Department of War's demands to use AI models for domestic mass surveillance and autonomous killing without human oversight.

there are hundreds of employees at these labs who are watching the administration target one of them. this should be a catalyzing moment notdivided.org

more broadly, we should seize on this moment: there is real opportunity to get policymakers on board for a no AI use in mass domestic surveillance or autonomous lethal weapons. now is the time.

It would take the Pentagon months to replace Anthropic’s AI tools: sources AI maker digs in with Thursday statement rejecting DOD pleas for unfettered use.

i am curious how this is supposed to play out in the next few weeks:

• OpenAI + Google may/may not have fights in signing up to replace Anthropic at DOD
• if one lab tries to articulate a similar red line, DOD may well try to label them a supply chain risk
• ???

www.defenseone.com/threats/2026...

The DoD fight is about much more than Anthropic Are Google and OpenAI prepared to enable mass state surveillance?

www.transformernews.ai/p/the-dod-fi...

But more important than what Anthropic has not agreed to is what other AI companies appear willing to accept. xAI has agreed to an all lawful use clause, and according to under secretary of defense for research and engineering Emil Michael, OpenAI and Google have agreed in principle. In giving the DoD free rein to use their tools without restrictions, those companies could be enabling a dystopian nightmare. As the Snowden files revealed, mass surveillance is nothing new. But AI could take it to an entirely new level, creating a true digital panopticon.

while DoD threatening to use the Defense Production Act to force Anthropic to drop guardrails is a worrying escalation, "more important than what Anthropic has not agreed to is what other AI companies appear willing to accept [...] giving the DoD free rein to use their tools without restrictions."

• this is already too long, so /fin

• i'm grateful to work with a set of colleagues who have a wide variety of backgrounds, skills, training, and interests. being able to identify the kind of research that is best in service of an advocacy goal requires an interdisciplinary staff. it's key to our ethos and i think to success.

• despite over a decade in the field, i still feel there are so many things we don't know: research that is critical for sharpening our advocacy to ensure policy interventions best address immediate harms, while supporting movements building toward a just economic system that meets people’s needs

• I think there have probably been 3 or 4 different "Upturns" over my tenure. processes that work for 4 people won't work for 10, and policies that work for 6 people won't work for 12 remote. being adaptable and willing to experiment is important and the same goes for programmatic work.

• i feel really lucky to have landed at a place where i get to do exactly what i want to do. if you know people who are trying to figure out how to "get in the field," i am always happy to have a call to talk with them and understand potential opportunities to get in the work.

• focusing on the real-world consequences of AI systems on housing, jobs, healthcare, financial security, and other important aspects of our lives is what matters. the precise contours of how the tech works, how it shows up, and how it interacts with our political, social, and legal systems matters.

• two things are true at the same time: we are working in a time of great peril, where peoples' material conditions are being determined more than ever by oppressive uses of technology. we are also working in a time of great opportunity, where more people than ever are mobilized and want change.