AI is the lynchpin of the whole political project. It’s the container for the parallel governance they have promised each other. I truly think that they are surprised that people don’t like it…and that they can’t yet wholesale punish everyone for not liking them.

Shared governance would prevent these errors. "By a margin of 9-to-1, she says, those who stopped to cast their vote with a dry-erase marker said they did not. “The people who decide we’re going to give this much money to ChatGPT are not the people who interact with students every day,” she says."

Exactly this: Mythos inherently favors attackers. High FPs mean it can't be run in continuous integration; defenders can only use it as part of an occasional audit. Attackers don't care about CI or development friction. They only need one exploit in an entire stack, and can wade through many FPs.

"What makes it possible for a [..] dictatorship to rule is that people are not informed; how can you have an opinion if you are not informed? If everybody always lies to you, the consequence is not that you believe the lies, but rather that nobody believes anything any longer." -- Hannah Arendt

Not just lobbying! our new paper maps 27 mechanisms of regulatory capture used by Big AI + 11 narrative framings that rationalise capture. “Big AI’s Regulatory Capture: Mapping Industry Interference and Government Complicity” will be presented at #FAccT2026 next month arxiv.org/abs/2605.068... 1/

We couldn’t agree more. At UT Austin, 4 departments are being consolidated under intense political pressure & a punishing timeline, w/ devastating consequences for students, staff & faculty. African American Studies, Mexican/American Studies, Women’s Gender & Sexuality Studies, &American Studies. 😡😭

So many universities including my own employer are doing this at graduation ceremonies rather than have an actual human read names, which is like the laziest, most disrespectful thing a college can do to a graduating student.

the whole episode is out today: www.youtube.com/watch?v=5twr...

Amid students organizing at campuses across the country and commencement speakers lining up to get booed, with polls of students and the public increasingly sour, administrators are not beating the allegations that betraying shared governance and forcing adoption are acts of class solidarity.

After discovering a recent article about LLMs fabricating analyses from nonexistent data (arxiv.org/abs/2603.21687), I did an experiment with our new lab-sanctioned chatbot. Well, here’s the result (in case that’s not clear, I never actually uploaded any grades-intro-cs.xlsx file).

A common comment on my videos about LLM "hallucinations" is that this isn't as much of a problem because LLMs "cite their sources" now. This is a category error. Citing a source implies it was consulted. LLMs do not consult, even with RAG. A citation might be *included* but it's not *consulted*.

Great episode, including voicing the necessary critique that privilege and socialization of computer science has made our profession among the hardest to unionize or to engage in acts of solidarity. Yet our technical skills are needed to investigate, explain, and counter the ongoing capture & harms.

This guy... He makes > $1M per year as University of Utah president and has overseen massive consolidation at the UofU of everything from IT (which caused my group to leave) to development, and even landscaping. He also likely worked with the Utah legislature to remove the power of the UofU Senate.

Well, it happened. A client disregarded my advice about using generative AI and guess what? It gave them bad advice AND the judge ruled that by using it, Client waived confidentiality and ordered us to turn a bunch of private stuff over. STOP USING GENERATIVE AI FOR LEGAL THINGS. I BEG YOU.

Hell yeah

Forget the “Who delivered the best commencement speech?” debate. Instead, give me scenes of students loudly and proudly denouncing speakers spreading AI hype.

This is legitimately one of the most horrifying things I have ever seen in all of the LLM bullshit that's out there. (That "joyful" is misspelled seems especially meaningful.)

New paper from @aial.ie! @harshp.com, Dick Blankvoort, Adel Shaaban, @sashamtl.bsky.social & me We analysed 6 GenAI ToS--finding missing info, major power imbalances & user obligations that are impossible to meet without violating the terms arxiv.org/abs/2603.18964 & aial.ie/research/ter... 1/

Issue 6: If an input violates the terms or acceptable uses, the user is solely liable. However, if an output violates the terms, the user is still solely liable! Apparently the provider can never be at fault. Even when they use automated black box filtering/detection invisible to users. 11/

When you are confident that you do a good job reviewing and cross-checking, that you can trust your colleagues and it won't happen to you, remember that this is how lawyers giving sworn testimony after being caught filing AI-generated citation misconduct describe the policy and process that failed.

Universities' "top-down implementation ignores severe institutional risks: imminent workforce displacement, inherent, unfixable data hallucinations, work intensification, systematic intellectual property theft, and widespread student skepticism."

It's interesting to compare the "Lesson Intro" with the AI-generated summary. The intro is communicating much more than the summary—namely, the instructor is setting a very specific tone with their students, helping show what kind of instructor they are and what kind of students they should be.

UNC Chappel Hill Trustees don't always interfere with tenured faculty hiring, but when they do, it's always rooted in fear of highly-qualified women of color.

Any sufficiently-advanced "AI" is indistinguishable from a cryptocurrency scam.

"You can't be having all these trans women being murdered by men in the headlines when you're trying to convince the public that the safest thing would be for trans women to use the men's bathroom and locker room. "You run the risk of people seeing through the propaganda that way."

The compounding tragedy of this is that university administrators will see this as a reason to lean in even more into commercialized research rather than basic research

Using an LLM in any professional capacity is in a sense an attempt to give oneself license to make "mistakes" without personal culpability. We cannot allow "the LLM made a mistake" to be an excuse in any human endeavor we actually give a damn about

The chatbot of the police in the Netherlands reacts to a report of incitement of violence against Palestinians as a case of antisemitism. Likely because the word Palestinian was mentioned. Thinking of the scholars who claimed recently AI can help democracies & governance. LOLwelp #ResistAI #Racism

Fake-citation cryptogyny. en.wikipedia.org/wiki/Cryptog... Also, fake citations are a temporary problem (tools exist to check, and also to insert fraudulent citations to real papers), but citation fraud (misrepresenting the substance of real papers) will remain and is labor-intensive to identify.

I was talking to someone about how the Canvas hack reminded me of Crowdstrike, in the sense that everyone going with the same thing that promises to do everything so that no one can be fired for buying it creates a dangerous monoculture. en.wikipedia.org/wiki/2024_Cr...

Poured through 100s of pages of filings to follow the power lines in the Lake Tahoe region, sitting in geographical and jurisdictional ambiguity. Now the 49k residents are in between California electricity rates and Nevada power sources. fortune.com/2026/05/12/l...

Feeling overwhelmed with all of this. LISTEN. Another Louisiana resident (Marshawn is the only name we have) testifies against the state’s racial gerrymander, speaking truth to power up and down the hearing room and predicts how this story will end. Speak your truth, fight and believe.

I desire more in life than to be the “don’t summarise with AI”-guy on bluesky 😅 but reports like this keep coming in, so I’ll repeat ad nauseam, don’t summarise with ‘AI’! www.tue.nl/en/our-unive...

📖 “The mechanism based limitations in the scope of applicability of LLMs are often masked in current discourse (…) This risks creating the illusion that LLMs can do things that they cannot, and that they have a connection to truth and understanding that they do not.” — Nicole Cruz (2026). 3/🧵

Gillibrand is referencing this article, published on April 9, and the article explains why "22 schools and 17 health care facilities" was likely an underestimate at that time. www.nytimes.com/interactive/...

I’ve heard anecdotally of several faculty planning to figure out Canvas alternatives this summer. Is there a group of folks and resources we could bring together? Is there something existing already?

Published with some minor changes! Please share widely. 😊 Guest, O., Suarez, M., ... & van Rooij, I. (2026). Against the Uncritical Adoption of 'AI' Technologies in Academia. Digital Culture & Education, 16(2), 85–118. doi.org/10.5281/zeno... Journal: www.digitalcultureandeducation.com/volume-162

sorry, i work at a university in a red state, and it’s illegal for the women & gender studies faculty to spend research money to attend their conferences. what in the fuck are they talking about

We can refuse this dystopia. What this moment requires is a radical rejection of the transactionalism that has infected higher ed. We need to show, through actions, that we embrace mutual trust and respect, friction, and genuine interaction. Ungrading can be a start. Doing so is scary, but necessary

Here's who Bernie seems to be taking "AI Safety" advice from, having become fully captured by the TESCREAL bundle and going full on doomer. Do you recognize the dudes that Max Tegmark is with in this so-called "AI Safety" summit? I'll help, it's Muskrat and Netanyahu.

Some seem to be taking the wrong lessons from this scene. kotaku.com/university-c... (1) It is 0% shocking that new college graduates (in humanities!) are critical of AI. (2) Pushing back on AI hype from a commencement speaker seems more likely an indication of high rather than low AI literacy.

The business model is deception.

"AI" is a culture and ideology. It doesn't matter what you thought when you put up the flag, that's what it means now. Developers who reject the ideology can simply not brand their technology as "AI", and then we can have a reasoned and nuanced conversation about the ethics of that technology.

Journalists, let’s be careful about the language we use when writing or talking about AI. Here’s a thread — @emilymbender.bsky.social, I know you pay attention to this, too, so feel free to chime in. (1/x)

folx, I’ve been working on AI "ethics" for almost a decade now & hate to tell you there’s no such thing as "ethical AI" the only "ethical principle" that works is regulating the shit out of the industry & to take back control of our autonomy, fundamental rights, freedoms & right to exist w/out AI

The errors and biases are serious unsolved problems, mostly unexamined as companies and medical executives race to adopt technology that would fail basic audits. Unions are the only institutions standing up. When the study can be done, we're going to find that unionization reduces clinical errors.

Missed Ghost in the Machine at SIFF? You can catch it via Kinema! And, through tomorrow, participate in a fundraiser for @dairinstitute.bsky.social while you're at it! notaidoc.com/dairxghost

How long until academic journals and conferences make it here?