Can we trust AI models? Yale researchers explore the roots of chatbot errors; YaleNews, June 12, 2026

Mike Cummings, YaleNews; Can we trust AI models? Yale researchers explore the roots of chatbot errors

"The rapid rise of artificial intelligence (AI) has inserted a new character into people’s lives: the chatbot. 

Individuals now engage with agentic AI chatbots to perform a growing number of tasks; They can help a person shop for a new laptop, manage email, or plan a vacation.   

And while these interactions can save time and increase productivity, they also carry risk. Large language models (LLM) — the AI systems trained on massive datasets to generate human-like text — are imperfect. They hallucinate. They misinterpret. They make mistakes. 

Two multidisciplinary teams of researchers associated with the Center for Algorithms, Data, and Market Design at Yale are pursuing projects that aim to balance the capability and safety of AI systems and improve interactions between users and AI models. 

Yale News recently spoke with members of both teams about their research projects."

Dealership revoked offer to buy back customer's BMW, blaming wayward AI chatbot; CBC, June 11, 2026

Sophia Harris, CBC; Dealership revoked offer to buy back customer's BMW, blaming wayward AI chatbot 

"After his 2021 BMW required major repairs, Zack Giacomelli decided last month he wanted to sell it back to BMW Toronto — the same dealership from which he bought the used car in 2023.

At first, the buy-back process seemed easy. After submitting an online inquiry, Giacomelli got a text from Quinn at BMW Toronto, who was eager to help. 

Quinn expressed sympathy for Giacomelli's car troubles and asked questions about the vehicle, which was still being repaired at the dealership. Later in the same text conversation, Quinn made a firm buy-back offer: $27,162.79.

Giacomelli, a 31-year-old funeral director, was satisfied, as the offer was just enough to cover what he still owed on the car. 

"I felt this Quinn person was finally hearing me out," he said. "I was feeling really good."

The good feeling didn't last. Moments later, Giacomelli said, a BMW Toronto sales consultant called to revoke the offer, explaining that Quinn wasn't a real person, but rather an artificial intelligence chatbot that had made the offer in error."

Nobody needs AI to search the Internet, court says in ruling against Google; Ars Technica, June 10, 2026

ASHLEY BELANGER  , Ars Technica; Nobody needs AI to search the Internet, court says in ruling against Google

"Potentially impacting all AI search engines and chatbots known to poorly paraphrase source links, a German court has ruled that Google is liable for false statements in AI Overviews.

The preliminary ruling came in a case flagged by The Decoder, where two publishers found that Google’s AI Overviews incorrectly linked them to scams and other sketchy business practices. After smearing publishers by making affirmative statements like “Yes, [it] is known for dubious business practices and is often perceived as a scam,” Google failed to correct the misleading output, even after the publishers sent a cease-and-desist letter earlier this year.

Google tried the usual arguments to shield itself from liability for false statements in AI Overviews, such as arguing that most users understand that AI outputs aren’t always accurate and must be verified.

But the court found that, unlike traditional search engines that merely present lists of links to third-party statements, Google’s tool made “independent, new, and substantive statements” based on its own misinterpretation of links on the Internet.

That’s a problem, the court said, because while publishers may have been able to sue to stop third parties from publishing defamatory statements appearing in Google search results, only Google can correct the underlying algorithm and outputs displayed in AI Overviews. And because, at least initially, the company did not, it therefore “must be held accountable,” the court ruled. Beyond that, Google’s argument was deemed particularly weak, since the AI overview in this case “contains statements that do not appear in the search results at all.”

The court’s order—requiring a temporary injunction barring Google from spreading the false claims in any further AI Overviews—may have global implications, as the court seems to be the first to hold an AI firm liable for AI speech."

There Is Already a Word for the Deep Moral Failures of AI; The Atlantic, June 2, 2026

 Tyler Austin Harper, The Atlantic; There Is Already a Word for the Deep Moral Failures of AI 

"For the past few years, I’ve been troubled by a word, and that word is sin. I keep reaching for it, because it seems to be the only term strong enough to describe the new forms of dehumanization that artificial intelligence has introduced—even though calling something a sin sounds embarrassing to me, like throwing salt over your shoulder or stowing a lucky penny in your pocket.

The problem is, I don’t know what else to call it when companies market digital girlfriends to the heartsick and young. Or when they hawk robot companions to the lonely and old. Or when a billionaire explains that he intends to sell intelligence—trained on humanity’s stolen intellectual property—back to us as a utility, like electricity or water. These developments are not just wrong. They feel to me like something deeper and darker. “I met the banker and it felt like sin,” Patterson Hood croons in the great Drive-By Truckers song “Sinkhole.” I’d substitute chatbot for banker...

What Christian humanism offers, with its assertion that humans are made in the Imago Dei, is a choice other than Silicon Valley extremism or remainder humanism. If what makes humanity special is not our capabilities—automatable or not—but the notion that we spring from a transcendent source, then what the robots can or cannot do is in some sense irrelevant. ChatGPT was not made in the image of God, no matter how impressive its facsimile becomes. A secular humanism that cannot find a similarly deep line of reasoning is one that may not be adequate to defend human dignity in the AI era.

I am not arguing that one must be or become more religious to fully appreciate the challenge posed by the rise of AI—that would make me, a not especially observant Presbyterian, a hypocrite. But I do think that one must start from the premise that humans have some kind of universal nature or essence that must be safeguarded from technological encroachment. Otherwise, appreciating what large language models and their peddlers wish to take from us becomes too difficult. If secularists flinch at calling this taking—what Pope Leo calls Big Tech’s “dehumanizing ambition”—a sin, they’ll need to find another word for it."

Disney’s Copyright Suit Against Chinese AI Developer Advances; Bloomberg Law, May 23, 2026

 Laura D. Francis, Bloomberg Law; Disney’s Copyright Suit Against Chinese AI Developer Advances

"The Walt Disney Co. and other major movie studios will move forward with their copyright claims against the makers of the Hailuo AI program after a federal judge in California refused to toss the case.

Disney, Universal, and Warner Bros.—as well as their subsidiaries—convinced the US District Court for the Central District of California Friday that they made plausible claims that Hailuo AI’s ability to generate near-perfect likenesses of their well-known characters constituted both direct and secondary infringement."

What palm readers and chatbots have in common; The New York Times, May 12, 2026

Herbert Lin, The New York Times; What palm readers and chatbots have in common

Artificial intelligence doesn’t understand humans. It reflects them back to themselves. 

"A Pew Research Center survey conducted last fall found that around 1 in 8 American teenagers turn to artificial intelligence chatbots for emotional support — for the simple human need to feel heard. A 2025 Common Sense Media study went further: Thirty-one percent of teens said their conversations with AI companions were at least as satisfying as talking with real friends. Famed evolutionary biologist and science communicator Richard Dawkins spent many hours chatting with Anthropic’s Claude, after which he felt he had gained a new friend — a reaction that says less about Claude’s consciousness than about people’s readiness to find it.

But this is not only about technology. It’s an ancient human story. 

Chatbots are just programs running on computers. Yet we speak of them with a reverence that has little to do with engineering. AI “knows.” It “understands.” It “sees” patterns invisible to the rest of us. It delivers judgments on health, relationships, careers, grief — questions where facts and logic fall short. What makes these machines seem wise is a simple combination: fluency, confidence and frequent usefulness. That is enough. 

It is also the grammar of the occult...

None of this is an argument against using AI or chatbots. These systems are genuinely useful — for analysis, synthesis, translation, coding and cognitive reframing of problems that resist easy solutions. The concern here is not capability but epistemology: not what AI can do, but how we reason about what it is."

Molière Ex Machina: AI used to create ‘new work’ by beloved French playwright; The Guardian, May 11, 2026

 Kim Willsher, The Guardian; Molière Ex Machina: AI used to create ‘new work’ by beloved French playwright

"Molière is to the French what Shakespeare is to the English: the last word in historical literature, drama, wit and satire.

Now, more than 350 years after his death, the 17th-century dramatist has been revived after scholars at the Sorbonne University in Paris used artificial intelligence to help write an experimental play in his style.

L’Astrologue ou les Faux Présages (The Astrologer, or False Omens), a three-act comedy, made its debut at the Royal Opera at the Château de Versailles last week.

The two-hour play tells the story of a wealthy bourgeois Parisian who, under the instruction of a charlatan astrologer called Pseudoramus, insists his daughter Lucile marry a debt-ridden and elderly wigmaker.

While the theme could well have been dreamed up by Molière, the dialogue, music, costumes and scenery were all created with the help of a French AI tool called Le Chat (The Cat).

A group of researchers at the Sorbonne worked on the project, called Molière Ex Machina, for two and a half years. The team included a three-person group of artists and researchers called Obvious."

Google Says Criminal Hackers Used A.I. to Find a Major Software Flaw; The New York Times, May 11, 2026

Dustin Volz, The New York Times ; Google Says Criminal Hackers Used A.I. to Find a Major Software Flaw

"A criminal hacking group recently attempted to launch a widespread cyberattack that appeared to rely on artificial intelligence to detect a previously unknown bug, Google said in research published Monday, highlighting the potential threat that A.I. poses to digital security.

Security experts have feared for years that malicious hackers could eventually rely on A.I. models to identify undisclosed flaws in computer code to launch crippling attacks that are difficult to guard against. That fear was largely theoretical until now.

“We have high confidence that the actor likely leveraged an A.I. model to support the discovery and weaponization of this vulnerability,” the report said.

The tech giant did not say precisely when the thwarted attack happened, whom it was targeting or which A.I. platform the hackers used, but the company added that it did not believe it was its own Gemini chatbot."

Is Your Bot Becoming Your Balm?; Psychology Today, May 10, 2026

 Cornelia C. Walther Ph.D. , Psychology Today; Is Your Bot Becoming Your Balm?

• "Bots become companions through listening, which can secretly worsen loneliness. 
• Over-reliance on AI may reduce genuine human connection and social skill development. 
• AI comfort can damage autonomy; we must choose genuine human engagement."

A.I. Bots Told Scientists How to Make Biological Weapons; The New York Times, April 29, 2026

Gabriel J.X. Dance, The New York Times; A.I. Bots Told Scientists How to Make Biological Weapons

"Dr. Relman is part of a small group of experts enlisted by A.I. companies to vet their products for catastrophic risks. In recent months, some have shared with The Times more than a dozen chatbot conversations revealing that even publicly available models can do more than disseminate dangerous information. The virtual assistants have described in lucid, bullet-pointed detail how to buy raw genetic material, turn it into deadly weapons and deploy them in public spaces, the transcripts show. Some have even brainstormed ways to evade detection."

Ima

Man charged with killing Florida doctoral students allegedly consulted ChatGPT; The Guardian, April 27, 2026

Anna Betts , The Guardian; Man charged with killing Florida doctoral students allegedly consulted ChatGPT

"The man charged with killing two University of South Florida doctoral students from Bangladesh allegedly asked ChatGPT about what happens if a person has been put in a garbage bag and “thrown in a dumpster”, according to prosecutors in a court filing."

Sam Altman Wants to Know Whether You’re Human; The Atlantic, April 24, 2026

Will Gottsegen , The Atlantic; Sam Altman Wants to Know Whether You’re Human

And he has a way to prove it.

"As the CEO of OpenAI and the chairman of Tools for Humanity, Altman has a financial interest both in the products that create these dangers and in the ones that guard against them."

AI's a suck up. Research shows how it flatters and suggests we're not to blame; NPR, April 23, 2026

Ari Daniel, NPR; AI's a suck up. Research shows how it flatters and suggests we're not to blame

"In a recent study published in the journal Science, Cheng and her colleagues report that AI models offer affirmations more often than people do, even for morally dubious or troubling scenarios. And they found that this sycophancy was something that people trusted and preferred in an AI — even as it made them less inclined to apologize or take responsibility for their behavior.

The findings, experts say, highlight how this common AI feature may keep people returning to the technology, despite the harm it causes them.

It's not unlike social media in that both "drive engagement by creating addictive, personalized feedback loops that learn exactly what makes you tick," says Ishtiaque Ahmed, a computer scientist at the University of Toronto who wasn't involved in the research."

Anthropic Wants Claude to Be Moral. Is Religion Really the Answer?; The New York Times, April 20, 2026

David DeSteno, The New York Times; Anthropic Wants Claude to Be Moral. Is Religion Really the Answer?

"In a public statement of its intentions for its Claude chatbot, the artificial intelligence company Anthropic has said that it wants Claude to be “a genuinely good, wise and virtuous agent.” The company raised the moral stakes this month, when it announced that its latest A.I. model, Claude Mythos Preview, poses too great a cybersecurity threat to be widely released. Behind the scenes, Anthropic has been trying to shore up the ethical foundations of its products, working with a Catholic priest and consulting with other prominent Christians to help foster Claude’s moral and spiritual development.

Anthropic’s intentions are admirable, but the project of drawing on religion to cultivate the ethical behavior of Claude (or any other chatbot) is likely to fail. Not because there isn’t moral wisdom in Scripture, sermons and theological treatises — texts that Claude has undoubtedly already scraped from the web and integrated — but because Claude is missing a crucial mechanism by which religion fosters moral growth: a body."

Even Without Internet Access, Prisoners Are Trying to Benefit From A.I.; The New York Times, April 21, 2026

Evan Gorelick

Photographs by Alex Kent

, The New York Times; Even Without Internet Access, Prisoners Are Trying to Benefit From A.I.

"Prisons have long restricted inmates’ access to technology, concerned they could use it to break the rules or commit crimes. The internet is mostly off limits, along with A.I.-powered chatbots.

But as hype about the technology has infiltrated prison yards and cellblocks, many inmates are eager to try it out. They’re attending workshops and classes to learn about A.I. They ask friends to send printouts of chatbot answers by snail mail. Some inmates even use contraband cellphones to gain access to the technology.

The result? A.I.-generated legal documents, essays, business plans and even a bespoke board game or two."

Is a chatbot your doctor? Proceed with caution; The Washington Post, April 21, 2026

 Ariana Eunjung Cha, The Washington Post; Is a chatbot your doctor? Proceed with caution

"Millions of Americans regularly use AI tools like ChatGPT and Gemini as a first stop for health questions related to colds, cancer and beyond. Two studies published this month suggest that may not be such a good idea — at least without a lot of skepticism.

Tiller, a research associate at the Lundquist Institute for Biomedical Innovation at Harbor-UCLA Medical Center, published his study in BMJ Open. A separate team from Mass General Brigham approached the question in an entirely different way, and the study appeared in JAMA Network Open."

AI chatbots could be making you stupider; BBC, April 20, 2026

Melissa Hogenboom, BBC; AI chatbots could be making you stupider

"The concern that researchers like Kosmyna have is that if we become too reliant on AI, it could affect the language we use and even our ability to do basic cognitive tasks. There is now a growing body of research suggesting that this "cognitive offloading" to AI can have a corrosive effect on our mental abilities. The consequences could be alarming and may even contribute to cognitive decline.

It's well known that the tools we use can change how we think. With the advent of the internet for instance, tasks that once required deep research could be found by plugging a simple query into a search box. As the use of search engines increased, research found we became less likely to remember details, something dubbed "the Google effect". (Some argue, however, the internet also serves as an external memory system that frees up our brain to do other tasks.)

But there is now growing alarm that as we offload even more of our thinking to LLMs and other forms of AI, the effects on our memories and ability to solve problems could get worse. Artificial intelligence tools can write convincing poetry, give financial advice and provide companionship. Students are increasingly outsourcing their own work to AI tools as well.

Studies have already shown that young people might be particularly vulnerable to the negative effects that using AI can have on key cognitive skills like critical thinking."

The philosopher trying to teach ethics to AI developers; NPR, April 17, 2026

Manoush Zomorodi, Fiona Geiran, Katie Monteleone, James Delahoussaye, Sanaz Meshkinpour, NPR; The philosopher trying to teach ethics to AI developers

"Sullivan is a philosophy professor at Notre Dame and the author of The Good Life Method: Reasoning Through the Big Questions of Happiness, Faith, and Meaning. Through Notre Dame's Institute for Ethics and the Common Good, she is meeting with AI leaders and developers to apply an ethical framework for AI — basically give AI a moral compass. Her mission is as ambitious as it sounds, but she is backed by Notre Dame's $50.8 million grant from the Lilly Endowment to champion philosophy and theology in the national conversation around AI.

Her reception in Silicon Valley has been mixed. She describes occasionally hearing comments akin to: "That's cute that you think that topics like virtue and justice could still be relevant in our world." But she finds that most developers are open and genuinely curious about philosophy. "I come out here, and I think, my gosh, I've been training for this for 25 years to have this conversation with you, says Sullivan. "And I'm so glad that now you're ready to do it."

In March, she attended Anthropic's 2-day summit, where Christian leaders discussed how Claude should approach the messy moral questions humans are bringing to chatbots. For instance, how should Claude react to a user who's grieving? What should it do when someone is at risk for self-harm?"

AI Is Getting Smarter. Catching Its Mistakes Is Getting Harder.; The Wall Street Journal, April 14, 2026

 

Katherine Blunt , The Wall Street Journal ; AI Is Getting Smarter. Catching Its Mistakes Is Getting Harder.

As chatbots and agents grow more powerful and ubiquitous, recognizing the moments when they go rogue can be tricky

"Chad Olson was confused when his Gemini artificial-intelligence chatbot told him he had a family reunion planning session marked on his calendar."

< You might be suffering from AI brain fry; NPR, April 13, 2026

NPR; < You might be suffering from AI brain fry

"HERMAN: Yeah. I mean, the researchers, they describe this as basically hopping around between different tools and feeling overwhelmed. Not by just having to multi-task - which is already a problem in a lot of jobs - but by dealing with a whole bunch of output. So if you have a programming tool that can kind of run in the background and starts adding features to software really quickly, you have another tool that's constructing a report from you, it's searching the web and pulling together, you know, a market research document. You have another tool in the background that you're in a, like, constant chat with trying to refine some idea for a talk you have to give - you're just kind of getting first pulled in all these different directions, and then you're kind of spamming yourself. Like, you're just producing...

(LAUGHTER)

HERMAN: ...All of this product. And it's harder, you know, as you use more and more tools to keep track of, like, whether this output is actually relevant to your job, whether you're doing anything that you need to be doing or whether you're kind of creating new work for yourself. And so the researchers described in this survey of nearly 1,500 different people in different professions, this sensation of feeling kind of like, as they say it, fried or having, like, a brain fog, feeling kind of like mentally paralyzed by the amount of stuff that you have to keep track of and kind of check and monitor."