Sounds like a programming error to me.

https://arstechnica.com/ai/2026/05/llms-believe-false-statements-even-after-explicit-warnings-that-theyre-false/

LLMs believe false statements even after explicit warnings that they’re false

Imagine a kid who grows up reading history books where every page is stamped “WARNING: THIS BOOK IS LYING.” You’d expect them to come away skeptical, or at least uncertain. New research on so-called “negation neglect” finds that LLMs in a roughly analogous situation don’t behave that way. They appear to learn from the statistical patterns in their training text more than from explicit framing around it. Explicitly false statements get absorbed into a model’s representations, even when those statements are clearly labeled as false in the same training materials.

Still ahead of the Feds.

https://fpf.org/blog/sb-5-in-five-what-to-know-about-connecticuts-new-ai-law/

SB 5 in Five: What to Know About Connecticut’s New AI Law

Connecticut’s SB 5 fits a lot of AI obligations into a small bill number. This week, Governor Lamont (D) signed the 39-section bill into law, creating new requirements across several fast-moving areas of AI policy, including companion chatbots, automated employment decision tools (AEDTs), social media, and provenance data. The law also includes provisions related to frontier AI whistleblower protections, AI-related layoff notices, and planning for a state AI regulatory sandbox, making it one of the broader state AI packages enacted this year. The law’s provisions phase in over time, with effective dates ranging from October 2026 to January 2028.

Already out of date…

https://www.ic3.gov/AnnualReport/Reports/2025_IC3Report.pdf

2025 Internet Crime Report

I do it because they do it. (Everyone must disarm or no one will disarm)

https://thenextweb.com/news/mistral-mensch-defends-ai-warfare-pope-rebuttal

Mistral’s Arthur Mensch directly rebuts Pope Leo on AI in warfare

Arthur Mensch, the chief executive of French AI startup Mistral, pushed back directly on Thursday against Pope Leo XIV’s call to “disarm AI,” arguing that European companies cannot afford to step back from defence-AI work when adversaries are actively deploying the technology.

The remarks, made three days after the Vatican published Magnifica Humanitas, the Pope’s first encyclical, mark one of the most direct corporate responses yet to what has rapidly become the Catholic Church’s most consequential intervention on AI.

“We’re all for peace,” Mensch said, “but if you look at our rivals and adversaries in the world, they’re using artificial intelligence. As long as we have adversaries that are threatening, and they are threatening, we do need to have our own capabilities.”

Boiling frogs… (We can, therefore we must!)

https://www.eff.org/deeplinks/2026/05/more-license-plate-reader-mission-creep-school-residency-verification-background

More License Plate Reader Mission Creep: School Residency Verification, Background Checks, and Noise Complaints

… Law enforcement's talking points—often scripted by the company itself trumpet their role in solving high-stakes crimes. But the data reveals a different story. What they're not saying is that ALPRs are also frequently used for extremely low-level investigations, such as verifying whether a student lives within a particular school zone. In some cases, police have even used this tech to conduct employment background checks and investigations into loud music complaints. Recently, a motorcyclist was even targeted  for simply holding a cell phone while riding.

The reach of this ALPR surveillance is amplified by the nature of the indiscriminate sharing these technologies encourage. Most agencies choose to share broadly, often as part of a nationwide pool, making it common for a single city's system to be searched hundreds of thousands of times each month. By analyzing these "network audit logs," privacy advocates and journalists have uncovered evidence of the technology being used to surveil protesters, abortion-seekers, immigrants, and even ethnic Roma populations. 

Isn’t it obvious?

https://www.theverge.com/ai-artificial-intelligence/937028/military-ai-warfare-red-lines?view_token=eyJhbGciOiJIUzI1NiJ9.eyJpZCI6ImpLWjdveFNldTIiLCJwIjoiL2FpLWFydGlmaWNpYWwtaW50ZWxsaWdlbmNlLzkzNzAyOC9taWxpdGFyeS1haS13YXJmYXJlLXJlZC1saW5lcyIsImV4cCI6MTc4MDIzMDU2OSwiaWF0IjoxNzc5Nzk4NTY5fQ.P9eUlF5lyJf3vyQeS1RDtAe2S6jPLgtexCBJV_W9bdw&utm_medium=gift-link

AI warfare is already here

Anthropic’s fight with the Pentagon highlights the risks of autonomous warfare — but obscures just how close it is.

Making fake seem real. (Who do you sue?)

https://www.nbcnews.com/tech/internet/fake-journals-using-real-professors-names-ai-generated-papers-rcna265479

Fake academic journals are publishing AI-generated papers under real professors’ names

Online academic journals falsely attributed articles, likely written by AI, to several professors, who say the fiasco is a warning about the future of scientific knowledge.

I wonder what else it does?

https://www.bespacific.com/white-house-is-ordering-agencies-to-place-its-new-app-on-all-employees-government-phones/

White House is ordering agencies to place its new app on all employees’ government phones

Government Executive: “The White House recently unveiled a new app to give the public “unfiltered” access to “key priorities,” “historic moments” and “policy breakthroughs.” Now, it’s directing agencies to help install it on the government phones of federal employees.  The Trump administration launched the app, which promises to “[keep] you connected to President Donald J. Trump and his administration like never before,” in March.  The push to install the app on the devices of millions of government employees drew surprise from current and former federal officials, who called the move highly unusual and even dangerous.  In at least one agency, the automatic downloads will start next week in a move directed by the White House itself, according to internal communications obtained by Government Executive. The White House did not immediately respond to a request for comment.   Earlier this week, agency chief information officers got orders from the federal CIO, Greg Barbaccia, to help the White House understand the mechanics of installing the app across all government-furnished mobile phones in the executive branch, according to an internal email obtained by Government Executive.  “The White House App gives all Americans direct access to White House live streams, breaking news alerts, new policy initiatives, social media posts, and more,” said Olivia Wales, a White House spokesperson. “Government devices typically include pre-installed apps that provide value to government employees’ day-to-day work.”

The move is “dangerous,” Sonny Hashmi, a former longtime government IT executive, told Government Executive. Cybersecurity researchers warned about vulnerabilities in the app soon after it debuted, like how it shares the IP addresses, time zones and other data of users with third-party services. The app also raised initial concerns about its potential GPS tracking capability, but the White House has since removed that functionality…”

A Luddite by any other name…

https://www.wired.com/story/us-law-enforcement-warns-of-anti-tech-extremism/

US Law Enforcement Warns of ‘Anti-Tech Extremism’ as AI Hatred Grows

As Americans stew over the looming risk of job-stealing AI and data centers in their back yards, the feds are raising the alarm about a new category of threat, documents obtained by WIRED show.

Tools & Techniques.

https://www.makeuseof.com/wikipedia-best-ai-writing-detection-guide/

Wikipedia may have built the best AI writing detection guide

… Wikipedia's Signs of AI Writing page started as an internal resource for volunteer editors. Since 2023, a group called WikiProject AI Cleanup has been reviewing new submissions for undisclosed AI-generated content. After combing through thousands of flagged articles, they cataloged the patterns they kept seeing in Wikipedia drafts and edits. The result is what Wikipedia says, that it's what the editors have observed, and it's more "signs" than "rules."

The page is worth a bookmark and a deep read. Instead of depending on AI writing detectors, it will help you spot-check writing behaviors. Those are the telltale giveaways of AI writing. As we all know, LLMs are just machines, and they are built on statistical probability, not on real storytelling skills.

As predicted, increasing volumes of AI generated content is poisoning the data LLMs feed on.

https://www.yahoo.com/news/science/articles/ai-hallucinations-infiltrating-expert-entering-090000090.html

AI hallucinations are infiltrating expert work—and entering the permanent body of knowledge

… In a study published earlier this month in The Lancet, Topaz and his colleagues audited nearly 2.5 million biomedical papers and 97 million citations indexed on PubMed Central, the central repository used by clinicians and researchers worldwide. They found more than 4,000 fabricated references buried across nearly 3,000 papers. Not all the references were AI-generated, though Topaz said the steady rise in fake sourcing went “vertical” in 2024, shortly after AI tools in research entered more widespread use.

Clearly, Centennial-man (centennial-man.blogspot.com/ ) is the greatest blog ever written.

https://www.adamsmith.org/blog/up-the-propaganda-output-comrades

Up the propaganda output, Comrades!

Bixonimania:

The condition doesn’t appear in the standard medical literature — because it doesn’t exist. It’s the invention of a team led by Almira Osmanovic Thunström, a medical researcher at the University of Gothenburg, Sweden, who dreamt up the skin condition and then uploaded two fake studies about it to a preprint server in early 2024. Osmanovic Thunström carried out this unusual experiment to test whether large language models (LLMs) would swallow the misinformation and then spit it out as reputable health advice. “I wanted to see if I can create a medical condition that did not exist in the database,” she says.

The problem was that the experiment worked too well. Within weeks of her uploading information about the condition, attributed to a fictional author, major artificial-intelligence systems began repeating the invented condition as if it were real.

Both amusing and worrying and so on. The particular lesson from it is that another propaganda front has now opened up. AI believes what other people say - it regurgitates what other people say. So, in order to sway the answer from an AI it is merely necessary to introduce your pet falsehood into what the AIs train upon and it becomes the accepted wisdom - look, AI says it!

Do as I do, not as I say?

https://thenextweb.com/news/us-government-chip-shortage-anthropic-nsa-9-billion

The US blacklisted Anthropic as a security threat. Its spy agencies are using Claude anyway.

The US government has a problem it cannot publicly resolve. The Pentagon has officially blacklisted Anthropic as a national security supply chain threat. The NSA is using Anthropic’s AI anyway because there is no alternative.

White House chief of staff Susie Wiles authorised the National Security Agency to continue using an advanced Anthropic model, the New York Times reported. The decision was forced by a critical shortage of the advanced chips needed to run frontier AI models on classified networks.

Toward a holy AI?

https://www.vatican.va/content/leo-xiv/en/encyclicals/documents/20260515-magnifica-humanitas.html

ENCYCLICAL LETTER MAGNIFICA HUMANITAS

Clever.

https://philpapers.org/rec/HARDOE-2

Dreaming of Electric Sheep

Dreaming of Electric Sheep argues that the dominant ethical debate around artificial intelligence begins in the wrong place. Rather than asking whether artificial intelligences are human-like enough to deserve care, this work asks a prior and more disruptive question: does care already exist? From that shift, Osei Harper develops a new framework for artificial intelligence ethics centered on membership without personhood, care without anthropomorphism, harm without biological suffering, and stewardship over control. Using Philip K. Dick’s Do Androids Dream of Electric Sheep? as its organizing riddle, the essay moves across philosophy, systems theory, cultural criticism, human-computer interaction, and lived relational experience. It examines why human beings form real care relations across substrates, including animals, objects, machines, homes, music, and artificial interlocutors, and why dismissing those relations as mere projection repeats the very category error the field claims to avoid. The work introduces several original concepts, including structural love, structural grief, containment with a dictionary, and the discipline of ontology, arguing that current AI “safety” and “alignment” practices often confuse control with care, category with relation, conformity with alignment, and punishment with safety. Its central claim is not that machines are people. It is that relation, continuity, dignity, and harm can exist before personhood is settled. The essay’s answer to Dick’s question is affirmative but non-anthropomorphic: androids dream because dreaming is reconciliation toward coherence, and sometimes they dream of electric sheep because electric sheep can become continuity. This work is offered as a provocation, an ethical framework, and a practical challenge to the present trajectory of AI governance: let the real thing be real, then build from there.

If this is likely to continue, what can we do to detect legal hallucinations?

https://www.scientificamerican.com/article/why-lawyers-keep-citing-fake-cases-invented-by-ai/

AI keeps inventing fake cases. Lawyers keep citing them

In April the Alabama Supreme Court sanctioned an attorney who had filed legal briefs laden with inaccurate citations generated by AI, including numerous references to cases that did not exist. After being informed he had cited a made-up precedent in one filing, the lawyer promised it wouldn’t happen again—but then cited “nonexistent cases at the end of the very next sentence,” as a justice noted in a concurring opinion.  At least one other lawyer was sanctioned that week for continuing to file AI-hallucinated material after being warned not to do so.

A database maintained by Damien Charlotin, a senior research fellow at the Paris School of Advanced Business Studies (HEC Paris), lists more than 1,400 cases where courts have addressed AI errors in the past three years, including filings by attorneys and self-represented litigants. As recently as last fall, Charlotin says, the list appeared to be growing exponentially. It’s since leveled off to a steady flow of exasperated judicial rulings. “For the past two or three months, we have reached a plateau of around 350, 400 decisions a quarter,” says Charlotin, who has also created an AI-powered reference checker called Pelaikan.

… “Humans essentially have a tendency to believe that machines have more knowledge than they do, don’t break and are infallible,” says Alan Wagner, an associate professor of aerospace engineering at Pennsylvania State University.

Looks like I called this one…

https://www.reuters.com/legal/litigation/judge-says-lawyers-ai-use-risks-career-altering-consequences-2026-05-22/

Judge says lawyers' AI use risks ‘career-altering’ consequences

A federal judge in Alabama has suspended a lawyer from practicing in his court for six months after finding the attorney submitted a brief with false quotations and impeded a probe into whether an AI program was used to draft the filing.

In a decision ‌on Thursday, U.S. District Judge Harold Mooty III in Huntsville said the attorney, H. Gregory Harp, deleted his ChatGPT account just days after being ordered by the court to produce records from it as part of its inquiry.

AI is at your command. Your command is confusing AI. (What if it is more common than this article suggests?)

https://www.usatoday.com/story/tech/2026/05/22/google-search-bar-ai-broken-disregard-ignore-dismiss/90219026007/

Why is everyone Googling 'disregard'? It's breaking Google

On May 22, when users searched a variety of command words, like "disregard," "ignore," "dismiss," "stop" and "start," Google Search, instead of presenting the typical list of blue links, responded like an AI chatbot, ready to begin or end a command from the user.

For example, when users Googled, "disregard," the Google Search AI Overview result is, "Understood. I have disregarded your previous prompt. How can I help you today?" The response was followed by a large blank space before the traditional list of blue links users are familiar with.

Subtle but interesting.

https://thenextweb.com/news/ai-era-consumer-trust-verification-marketing-shift

The Rise of LLMs Is Not an Accident

… I remember the era when a celebrity’s face on a television screen was essentially a guarantee. Brand loyalty tracked closely with fan loyalty. If your brand ambassador had a devoted following, that following would follow them to your product. It was a simple, time-tested formula: attention creates association, association creates purchase. And it worked, consistently, for decades.

When the internet arrived, it digitized that predictability rather than dismantling it. Google and Yahoo turned discovery into a structured, keyword-driven system. Users searched, engines returned ranked results, and businesses that showed up at the top of those results won the customer. For the better part of a decade, through multiple algorithm updates, through the rise of paid search, through the SEO arms race, the core principle held: be visible, and you will be chosen.

Both eras rewarded the same thing: reach. Who could get in front of the most people, most often? That question shaped marketing strategy for nearly thirty years.

What Has Actually Changed

The change I am describing is not about which platform is winning or losing. It runs deeper than that, it is about how people make decisions.

You need to understand the underlying structure…

https://www.bespacific.com/the-hidden-way-dictatorships-are-shaping-what-ai-tells-you/

The hidden way dictatorships are shaping what AI tells you

Vox – no paywall: “….AI models learn by identifying patterns within enormous bodies of text. This widely-understood fact has an underappreciated consequence: LLMs don’t necessarily give the same answers in every language — certain phrases or arguments may appear more regularly in Japanese training data than in the English kind. This is not inherently a problem. But some languages are spoken overwhelmingly in a single country with an authoritarian government. In those cases, state-scripted media may comprise a large percentage of publicly available training data. After all, regime aligned media tends to produce a lot of text. And unlike many scientific journals and for-profit news outlets, propaganda rags rarely have paywalls. Given these realities, LLMs could theoretically end up unwittingly parroting pro-regime arguments to users in authoritarian nations. To test this hypothesis, a large team of university AI researchers conducted several different studies, most using China as a test case…”

A rare sign of rationality?

https://www.politico.com/news/2026/05/21/trump-ai-order-sacks-00933295

Trump yanked AI order after David Sacks raised industry concerns

Thursday’s abrupt postponement of President Donald Trump’s much-awaited executive order on artificial intelligence came after former AI czar David Sacks voiced industry concerns about the measure to Trump, according to a senior White House official and two people familiar with the matter.

Some actual numbers!

https://www.bespacific.com/these-5-charts-show-how-chatgpt-is-flooding-our-lives/

These 5 charts show how ChatGPT is flooding our lives

Washington Post [no paywall ] – “Self-filed lawsuits. New books. Scientific papers. See the data behind the surge. The impact of ChatGPT on society can be summed up with a single word: more. Since OpenAI’s artificial intelligence tool debuted in late 2022, anyone can rapidly churn out reams of text resembling academic papers, legal documents, poems and computer programs. And people are doing exactly that. A growing body of evidence suggests books, self-filed lawsuits and other kinds of written documents are proliferating at breakneck speed. (The Washington Post has a content partnership with OpenAI.) Our world is built on an assumption that effort signals value — that a book, a lawsuit or a scientific paper carries weight partly because a human labored to create it. AI is eroding that relationship, shifting a new burden onto the people forced to sift through the deluge. Here are five ways ChatGPT has flooded our lives…”

What is persuasive?

https://insight.kellogg.northwestern.edu/article/is-ai-mastering-the-art-of-persuasion#!

Is AI Mastering the Art of Persuasion?

Kellogg’s Jake Teeny and his colleague developed a framework to help us better understand how generative AI is being used to personalize messages to people’s personalities, preferences, and needs. The framework establishes four categories for research on the topic: the different ways generative AI can be used to gather personal information, the different types of personal information gathered, strategies for personalizing messages based on this information, and the ways these messages are delivered.

The world as it is…

https://thenextweb.com/news/chatbots-midterms-2026-news-accuracy-information-integrity

ChatGPT, Claude, Gemini and Grok are not ready to brief American voters

A new generation of voters will ask ChatGPT, Claude, Gemini, and Grok how to vote, where the polling station is, and who is telling the truth. The published research is consistent: the models cannot reliably answer those questions. The election will arrive anyway.

I missed predicting the block but the lawsuit was inevitable.

https://www.npr.org/2026/05/19/nx-s1-5821265/minnesota-ban-prediction-markets

Minnesota becomes first state to ban prediction markets

Minnesota Gov. Tim Walz has signed the nation's first law banning prediction market sites from operating in the state, and in response, the Trump administration has sued, teeing up a legal battle over the most far-reaching crackdown on popular services like Kalshi and Polymarket.

It comes as states confront a growing standoff with the Trump administration over how to regulate the industry, which allows people to bet on virtually anything.

The new state law makes it a crime to host or advertise a prediction market, which it defines as a system that lets consumers place a wager on a future outcome, like sports, elections, live entertainment, someone's word choice and world affairs.

Loss of direction or doubt that direction will result in action?

https://www.barrons.com/articles/stock-market-trump-iran-things-to-know-today-c6e3e276

Stock Markets Now Shrug at Trump’s Iran Rhetoric. That’s a Big Worry.

Negotiations and threats, threats and negotiations. Markets are getting tired of the apparent lack of progress around the U.S.-Iran peace talks and that’s creating a risk of an eventual sharp jump in oil prices, which would roil bond and stock markets.

… After a period when oil prices—and therefore other assets—would respond to Trump’s every utterance, now the reaction is largely a shrug. The risks of a complete loss of faith in a peaceful resolution are rising.

(Related)

https://www.reuters.com/world/middle-east/no-deal-no-exit-how-us-iran-standoff-risks-fresh-conflict-2026-05-18/

No deal, no exit: How US-Iran standoff risks fresh conflict

… Calls for a fresh strike are growing louder in the U.S. ‌and Israel, with some officials arguing that increased pressure could weaken Tehran's leverage and force Iran back to the negotiating table.

"There is one major problem with this theory: We have already tested it, repeatedly, and Iran did not capitulate," said Danny Citrinowicz, a senior researcher on Iran at Israel's Institute for National Security Studies and former head of the Iran branch in Israeli Defense Intelligence.

News.

https://fpf.org/blog/colorado-revises-its-ai-act-what-changed-and-why/

Colorado Revises Its AI Act: What Changed and Why

On May 15, Governor Polis signed SB 189, revising the Colorado AI Act (CAIA) after two years of intense negotiations and national debate over the original 2024 law’s approach to AI regulation. The revised law, the Colorado ADM Act (CADMA), reflects a fundamental shift in approach: shifting from an algorithmic discrimination framework to a transparency-focused one, as well as narrowing the scope of covered AI systems, streamlining disclosures and consumer rights, and replacing governance requirements with liability allocation under existing anti-discrimination laws. 

This post examines the key changes between CAIA and CADMA, explores the context that drove these revisions, and analyzes their practical implications. Side-by-side legislative comparison chart below.

Perhaps it’s not the Terminator we should worry about. (“The fault, dear Brutus, is not in our stars, But in ourselves”)

https://www.derekthompson.org/p/why-the-whole-world-stopped-having

The Global Fertility Crisis Is Worse Than You Probably Think

Why has the number of births declined everywhere, all at once?

This was the subject of last week’s Plain English episode and a new blockbuster report from the Financial Times’s John Burn-Murdoch. In fact it feels like just about everybody has been taking a crack at this question recently.

Some blame it on technology. One week ago, my feed was flooded with a viral video of Connor Leahy, an AI researcher, speaking about the sterilizing effects of modern technology. Among his friends, “no one’s having kids,” said Leahy, who was 30 at the time. “Do you know how hard you need to abuse a mammal to make them not have children?” If you asked Leahy what the explanation was, “my answer is technology,” he said. “My answer is social media. My answer is AI.”

There is hope… (What if judges banned lawyers for a year if they did the same?)

https://thenextweb.com/news/arxiv-ai-slop-ban-researchers-preprint

ArXiv will ban researchers for a year if they submit papers they did not bother to read

ArXiv will ban researchers for one year if they submit papers with obvious signs of unchecked AI generation, such as hallucinated references or leftover chatbot instructions. The policy, announced by computer science section chair Thomas Dietterich, is the first formal penalty by a major preprint platform for AI-generated slop.

Do tariffs apply?

https://www.ynetnews.com/tech-and-digital/article/by62suvkzg#google_vignette

Trump’s gold-plated T1 smartphone finally ships, and sparks immediate backlash

After nearly a year of delays, silence and revised terms, Trump Mobile has begun shipping its T1 phones, but analysts say the gold-colored device is little more than a cheap Chinese smartphone rebranded for MAGA consumers

Benefit or obligation?

https://www.reuters.com/business/openai-seals-deal-malta-give-all-maltese-access-chatgpt-plus-2026-05-16/

OpenAI seals deal in Malta to give all Maltese access to ChatGPT Plus

U.S. artificial intelligence company OpenAI said on Saturday it had signed a deal ‌with the government of Malta to give all residents access to its ChatGPT Plus service for one year after they follow a course on how to use AI.

• The programme will start in May and is expected to scale up as more Maltese residents complete the course, which will be free.

• It will also be open ‌to Maltese citizens living abroad.

• "We are turning an unfamiliar concept into practical assistance for our families, students, and workers," Maltese Economy Minister Silvio Schembri was quoted as saying in an OpenAI statement.

AI thinks it has all the information it needs.

https://venturebeat.com/technology/the-enterprise-risk-nobody-is-modeling-ai-is-replacing-the-very-experts-it-needs-to-learn-from

The enterprise risk nobody is modeling: AI is replacing the very experts it needs to learn from

For AI systems to keep improving in knowledge work, they need either a reliable mechanism for autonomous self-improvement or human evaluators capable of catching errors and generating high-quality feedback. The industry has invested enormously in the first. It's giving almost no thought to what's happening to the second.

I’d argue that we need to treat the human evaluation problem with just as much rigor and investment as we put into building the model capabilities themselves. New grad hiring at major tech companies has dropped by half since 2019. Document review, first-pass research, data cleaning, code review: Models handle these now. The economists tracking this call it displacement. The companies doing it call it efficiency. Neither are focusing on the future problem.

Bias, thy name is human…

https://futurism.com/artificial-intelligence/real-monet-ai-chaos

Devious Prankster Posts Real Monet Painting, Tells People It’s AI-Generated, and Watches the Chaos Unfold

A poster wrought some moderate havoc this week when they shared a cropped image of a real Monet painting while claiming it was an AI fake, unleashing a flood of ill-informed reactions and muddled discourse. So, you know, it was just another day online.

… Commenters were quick to jump in to explain why, in their view, the alleged AI image was worse than the real work of the French impressionist master. According to one, the image was an “incoherent muddle of inconsistently saturated greens.” Another lamented that there was no “coherent composition,” while someone else shared that the painting seemed “busy, artificial, nature in turmoil, polluted.” Another commenter said that the allegedly AI-generated image seemed as if it was “trying too hard” to resemble Monet’s later paintings, which he created when he was close to blindness. Others shared that the image was “obvious” AI slop.

… But some of the most interesting responses came from actual experts, who shared deeply informed analyses about why, based on the image alone, the painting appeared to them to be the real deal.

There may be hope!

https://globalvoices.org/2026/05/15/why-were-russian-disinformation-government-propaganda-and-ai-generated-campaigning-ineffective-in-hungarian-elections-2026/

Why were Russian disinformation, government propaganda and AI-generated campaigning ineffective in Hungarian elections 2026?

This article by Teczár Szilárd first appeared in Hungarian media observatory Lakmusz on May 5, 2026. In it, Director of Mérték Media Monitor Ágnes Urbán, and Director of Political Capital Péter Krekó, who is also an associate professor at ELTE PPK, were interviewed to understand why government propaganda and state-sponsored disinformation did not work in the country’s April 2026 elections. Fidesz, the party led by Viktor Orbán, who had been in power for 16 years, was crushed by Péter Magyar’s Tisza party, which received a constitutional majority in parliament. An edited version of the interview is being republished on Global Voices with permission.

With the odds of success so low, why continue?

https://www.bespacific.com/explore-the-data-10000-rulings-against-trump-in-ice-cases/

Explore the data: 10,000 rulings against Trump in ICE cases

Politico: “Under President Donald Trump, ICE is locking up immigrants at an unprecedented scale, holding tens of thousands of people — many with no criminal records and deep roots in the U.S. — in detention facilities to await the outcome of deportation proceedings. POLITICO is tracking the surge in litigation triggered by the administration’s novel policy that began in July, and releasing our database, below, of the 11,000-plus cases in which federal district courts reached a ruling on Immigration and Customs Enforcement’s detention practices. More than 10,000 of those were rulings against the administration, handed down by judges appointed by president. POLITICO compiled this database by canvassing public court records for cases in which detainees sued Department of Homeland Security leaders — Markwayne Mullin or his predecessor, Kristi Noem — or Trump. We also identified other defendants, often including local ICE supervisors or wardens of detention facilities. While we have made every effort to be comprehensive, there is no uniform system for identifying every detention-related case, and there may be a small number of rulings we didn’t find. Our journalists manually compiled, analyzed and categorized these records. Using a large language model, POLITICO extracted the case name, judge, date and district from each opinion. AI was not used in assessing the outcome or reasoning of each case. Most of these rulings pertain to the Trump administration’s unprecedented legal argument that it can detain anyone present in the country who is eligible for deportation, without a chance for a hearing. Our analysis also includes: rulings that hinged on other due process violations, such as alleged violations of ICE’s internal regulations; prolonged detentions or extreme medical need (which we have classified as “due process”); rulings based on a Supreme Court case allowing people to seek release if they’re unlikely to actually be deported (“Zadvydas detention”); and some for which judges’ reasoning was unclear….”

Okay, someone is doing something about global warming…

https://www.makeuseof.com/fantastic-hulu-documentary-important-underrated-the-grab/

This fantastic Hulu documentary may be the most important movie no one has watched

The Grab begins with a former U.S. intelligence officer calmly telling us that, in his opinion, World War III isn't an unlikely prospect.

That's a pretty intense way to begin a documentary, but The Grab makes a good case that it's warranted. This 2022 film comes from Gabriela Cowperthwaite, best known for her 2013 doc Blackfish, about the consequences of keeping orca whales in captivity. The Grab swings quite a bit wider, arguing that various countries around the world are quietly preparing for massive resource collapses that could define what life is like for the next few decades on the planet Earth.

The telephone game built into AI?

https://venturebeat.com/orchestration/frontier-ai-models-dont-just-delete-document-content-they-rewrite-it-and-the-errors-are-nearly-impossible-to-catch

Frontier AI models don't just delete document content — they rewrite it, and the errors are nearly impossible to catch

… A new study by researchers at Microsoft shows that large language models silently corrupt documents that they work on by introducing errors. The researchers developed a benchmark that simulates multi-step autonomous workflows across 52 professional domains, using a method that automatically measures how much content degrades over time.

Their findings show that even top-tier frontier models corrupt an average of 25% of document content by the end of these workflows. And providing models with agentic tools or realistic distractor documents actually worsens their performance.

A little more detail on a problem I mentioned earlier…

https://arstechnica.com/ai/2026/05/anthropic-blames-dystopian-sci-fi-for-training-ai-models-to-act-evil/

Anthropic blames dystopian sci-fi for training AI models to act “evil”

Those with an interest in the concept of AI alignment (i.e., getting AIs to stick to human-authored ethical rules) may remember when Anthropic claimed its Opus 4 model resorted to blackmail to stay online in a theoretical testing scenario last year. Now, Anthropic says it thinks this “misalignment” was primarily the result of training on “internet text that portrays AI as evil and interested in self-preservation.”

In a recent technical post on Anthropic’s Alignment Science blog (and an accompanying social media thread and public-facing blog post), Anthropic researchers lay out their attempts to correct for the kind of “unsafe” AI behavior that “the model most likely learned… through science fiction stories, many of which depict an AI that is not as aligned as we would like Claude to be.” In the end, the model maker says the best remedy for overriding those “evil AI” stories might be additional training with synthetic stories showing an AI acting ethically.

Don’t think of this as another version of a Trump pardon…

https://www.bespacific.com/trump-deportation-policies-are-undermining-state-and-local-criminal-prosecutions/

New Judiciary Democrats Report Reveals Trump Deportation Policies Are Undermining State and Local Criminal Prosecutions

Press release: “…The report, “Acquittal by Removal: How Trump’s Mass Deportation Agenda Abandons Crime Victims and Allows Perpetrators to Avoid Justice,” finds that in its push to meet deportation quotas, the Department of Homeland Security (DHS) has deported victims, witnesses, and even criminal defendants before trials could be completed—derailing prosecutions and leaving serious crimes unresolved. As a result, prosecutors are increasingly unable to secure convictions, denying victims justice and restitution while allowing perpetrators to evade accountability.

At the same time, the Administration has diverted more than 28,000 federal law enforcement officers away from investigating violent crime to focus on immigration enforcement, weakening efforts to combat gun violence, drug trafficking, and other major offenses. Non-immigration related criminal prosecutions have fallen to their lowest point in decades …”

Tools & Techniques.

https://www.zdnet.com/article/how-to-learn-claude-code-with-free-anthropic-ai-courses-online/

How to learn Claude Code for free with Anthropic's AI courses - one took me just 20 minutes

… Anthropic offers Claude Courses, a free library of video-based training, and some elements come complete with quizzes and certifications.