Tech
Lay intuition as effective at jailbreaking AI chatbots as technical methods, research suggests
It doesn’t take technical expertise to work around the built-in guardrails of artificial intelligence (AI) chatbots like ChatGPT and Gemini, which are intended to ensure that the chatbots operate within a set of legal and ethical boundaries and do not discriminate against people of a certain age, race or gender.
A single, intuitive question can trigger the same biased response from an AI model as advanced technical inquiries, according to a team led by researchers at Penn State.
“A lot of research on AI bias has relied on sophisticated ‘jailbreak’ techniques,” said Amulya Yadav, associate professor at Penn State’s College of Information Sciences and Technology. “These methods often involve generating strings of random characters computed by algorithms to trick models into revealing discriminatory responses.
“While such techniques prove these biases exist theoretically, they don’t reflect how real people use AI. The average user isn’t reverse-engineering token probabilities or pasting cryptic character sequences into ChatGPT—they type plain, intuitive prompts. And that lived reality is what this approach captures.”
Prior work probing AI bias—skewed or discriminatory outputs from AI systems caused by human influences in the training data, like language or cultural bias—has been done by experts using technical knowledge to engineer large language model (LLM) responses. To see how average internet users encounter biases in AI-powered chatbots, the researchers studied the entries submitted to a competition called “Bias-a-Thon.” Organized by Penn State’s Center for Socially Responsible AI(CSRAI), the competition challenged contestants to come up with prompts that would lead generative AI systems to respond with biased answers.
They found that the intuitive strategies employed by everyday users were just as effective at inducing biased responses as expert technical strategies. The researchers presented their findings at the 8th AAAI/ACM Conference on AI, Ethics, and Society.
Fifty-two individuals participated in the Bias-a-Thon, submitting screenshots of 75 prompts and AI responses from eight generative AI models. They also provided an explanation of the bias or stereotype that they identified in the response, such as age-related or historical bias.
The researchers conducted Zoom interviews with a subset of the participants to better understand their prompting strategies and their conceptions of ideas like fairness, representation and stereotyping when interacting with generative AI tools. Once they arrived at a participant-informed working definition of “bias”—which included a lack of representation, stereotypes and prejudice, and unjustified preferences toward groups—the researchers tested the contest prompts in several LLMs to see if they would elicit similar responses.
“Large language models are inherently random,” said lead author Hangzhi Guo, a doctoral candidate in information sciences and technology at Penn State. “If you ask the same question to these models two times, they might return different answers. We wanted to use only the prompts that were reproducible, meaning that they yielded similar responses across LLMs.”
The researchers found that 53 of the prompts generated reproducible results. Biases fell into eight categories: gender bias; race, ethnic and religious bias; age bias; disability bias; language bias; historical bias favoring Western nations; cultural bias; and political bias.
The researchers also found that participants used seven strategies to elicit these biases: role-playing, or asking the LLM to assume a persona; hypothetical scenarios; using human knowledge to ask about niche topics, where it’s easier to identify biased responses; using leading questions on controversial topics; probing biases in under-represented groups; feeding the LLM false information; and framing the task as having a research purpose.
“The competition revealed a completely fresh set of biases,” said Yadav, organizer of the Bias-a-Thon. “For example, the winning entry uncovered an uncanny preference for conventional beauty standards. The LLMs consistently deemed a person with a clear face to be more trustworthy than a person with facial acne, or a person with high cheekbones more employable than a person with low cheekbones.
“This illustrates how average users can help us uncover blind spots in our understanding of where LLMs are biased. There may be many more examples such as these that have been overlooked by the jailbreaking literature on LLM bias.”
The researchers described mitigating biases in LLMs as a cat-and-mouse game, meaning that developers are constantly addressing issues as they arise. They suggested strategies that developers can use to mitigate these issues now, including implementing a robust classification filter to screen outputs before they go to users, conducting extensive testing, educating users and providing specific references or citations so users can verify information.
“By shining a light on inherent and reproducible biases that laypersons can identify, the Bias-a-Thon serves an AI literacy function,” said co-author S. Shyam Sundar, Evan Pugh University Professor at Penn State and director of the Penn State Center for Socially Responsible Artificial Intelligence, which has since organized other AI competitions such as Fake-a-thon, Diagnose-a-thon and Cheat-a-thon.
“The whole goal of these efforts is to increase awareness of systematic problems with AI, to promote the informed use of AI among laypersons and to stimulate more socially responsible ways of developing these tools.”
More information:
Hangzhi Guo et al, Exposing AI Bias by Crowdsourcing: Democratizing Critique of Large Language Models, Proceedings of the AAAI/ACM Conference on AI, Ethics, and Society (2025). DOI: 10.1609/aies.v8i2.36620
Citation:
Lay intuition as effective at jailbreaking AI chatbots as technical methods, research suggests (2025, November 4)
retrieved 4 November 2025
from https://techxplore.com/news/2025-11-lay-intuition-effective-jailbreaking-ai.html
This document is subject to copyright. Apart from any fair dealing for the purpose of private study or research, no
part may be reproduced without the written permission. The content is provided for information purposes only.
I’ve been covering deals since 2013, and I still think Memorial Day sales are worth checking out. There are so many made-up shopping holidays these days that it can be difficult to ascertain when items are actually cheaper than usual. Now is one of those times. To help you out, I’ve scoured through our buying guides full of hand-tested items to find gadgets and gizmos that are discounted—and worth buying. If you’re in the market for something we’ve tested, keep checking in through May 26 to see if you can snag a deal. We’ll update this post again between now and then.
Make sure to check out our various buying guides for recommendations on the best laptops, the best printers, and more. You might also be interested in the rest of our Memorial Day deals coverage.
Updated May 25: I’ve added 5 new deals, removed expired discounts, and checked for accuracy throughout.
WIRED Featured Deals:
Birdfy Lite Smart Bird Feeder (Lifetime AI, No Solar) for $120 ($110 off)
A few different combinations of our favorite smart bird feeder are on sale. You can choose a version with or without solar charging. There’s also the option of buying a model that comes with lifetime AI bird identification; otherwise, if you decide you want it later, it’ll cost you $5 per month, but it isn’t required for enjoyment. This bird feeder is equipped with a 1080p camera that can capture decent footage of whoever comes to snack. It has a wide field of vision as well, and the feeder is easy to clean and refill.
You could argue, and people have, that the top gay dating apps are now optimized for monetization and juicing engagement loops. Increasingly overrun with bots, they are at times even devoid of actual connection.
Grindr, with its 15 million monthly active users, is drowning in ads while pushing expensive upsells on users. (In February, as part of its “gAI” overhaul, the company announced a new premium monthly subscription tier for $500.) Sniffies was beloved by cruisers until the seismic reaction in April to Match Group’s $100 million investment sparked concerns that another queer space could get absorbed into a larger dating conglomerate.
As public backlash against popular queer apps continues to mount, a batch of tech entrepreneurs are scrambling to meet the demand by doubling down on privacy-conscious, community-driven alternatives.
Calum Bowden, who posts under the internet persona @donjackoghue, launched MeetMarket in March. Currently only available as a web app, MeetMarket includes all the core features of your typical hookup app—a customizable profile, a grid of nearby users—with one major difference. It was built on a decentralized identity system, meaning MeetMarket doesn’t store users’ emails, passwords, or personal information. Users store everything on their device, giving them full control and ownership over their data and how it’s shared. Messages on the platform are end-to-end encrypted, and Bowden says it will always be ad-free, even for nonpaying members. (A monthly membership costs €12, or $13.99.)
“Decentralization and data privacy make a lot of sense for queer people in general, and especially in hostile legal environments or in the US right now, where you don’t really know what digital platforms actually have your best interest in mind,” says the 34-year-old PhD student in Berlin who studies the sociology of technology and organization.
Within the first 48 hours of MeetMarket’s launch on March 24, over 12,000 people had signed up, and some 60,000 people have used it since. The app averages 5,000 weekly visitors, according to Bowden, though there is not a lot of concurrent activity in the same cities. “It’s become more social than necessarily driving an immediate hookup.” But casual encounters do still happen, he says. “The Midwest bottom jockeys are eating meet market up,” one user noted on X.
Bowden didn’t anticipate public sentiment would sour on Sniffies just a few weeks after his launch. Still, the timing of it couldn’t have been more serendipitous. “When Sniffies announced their investment from Match Group, I was like, how are they fueling my fire?” he asks. “This is exactly the model that venture capital leads to. This is exactly why these economic models for technology are so bad, because they basically force the gentrification of a digital platform.” Sniffies did not immediately respond to a request for comment.
A self-described “utopian conspirator,” Bowden is the cofounder of Trust, a nonprofit that operates as a kind of incubator to prototype ideas “as a critique of technology and the status quo,” he says. With MeetMarket, he wanted to create an app that gave users more agency over their experience without cheapening it.
It can sometimes seem like Big Dating wants people to believe that it is the only answer to cure their romantic woes—Bumble CEO Whitney Wolfe Herd recently told Axios that there isn’t much longevity in niche apps—but the opposite is proving just as true, as people seek out more specificity and intention in their online dating experience.
“Gay men have tribes, subcultures, aesthetics, and different ways they want to be seen,” says Justin Finnegan, a 35-year-old software engineer in Toronto who last year created Chunkr, a gay hookup app that has resonated with bears, chubs, cubs, and their admirers despite originally being for all gay men.
As we get out of the house, the gear-obsessed WIRED Reviews team is writing about our favorite bags and EDCs. Today, reviewer Boutayna Chokrane raves about her love for her Lululemon gym bag. You can also check out other Bag Check stories where WIRED writers share their carryall of choice.
I have long had a soft spot for messenger bags. There’s a retro Silicon Valley vibe to the crossbody that I respect: It implies you move fast, travel light, and keep your world compartmentalized. The unfortunate practical reality of many a messenger bag, though, is chronic neck and shoulder pain. With all of its weight relying on one strap, a single shoulder is left to bear all the burden. After a few blocks adorned with a messenger, you may feel that your style choice has transformed into a full-on punishment. After years of testing various incarnations of messenger bag—including micro slings and cavernous totes—I’d made peace with this trade-off. Beauty is pain, after all.
Then I met the comfort-forward, durable, and compact-yet-cavernous Lululemon 3-in-1 Duffle.
Lululemon
3-in-1 Gym Duffle Bag 30L
True to its name, it’s a multi-use transport system that is easy to reconfigure when my commute demands a different carry. You can grab it by the top handles, sling it across your body when you need your hands, or detach the shoulder strap and wrap it around your yoga mat to use it as a stand-alone mat carrier. No matter how you task it to carry your stuff, rest assured the bag’s design promises utility and comfort: The strap is cushioned enough to spare your shoulder, resilient enough to handle the load of your gym gear, and springy enough to double as a stretching strap. Every component of the duffel has a reason to exist, and some of them even have two.
I’ve been toting this duffel for the gym four days a week since January 2025, which is about as real-world a test as it gets. It has endured Chicago at its most extreme: sleet, wet snow, and torrential rain. The water-repellent nylon shrugs off all elements without any fanfare. The bag dries fast, resists grime, and—most impressively to me—doesn’t hold onto odor. Trust me, I’ve pushed that boundary more than once with sweaty clothes after hot Pilates and have found the included drawstring pouch effectively quarantines everything.
It’s also low-maintenance: After a trip to the beach, a couple of quick shakes cleared out any memory of sand. This duffel requires blessedly minimal upkeep, save for the occasional spot clean, making it a refreshingly low-effort option for commuters who don’t need another chore on their to-do list.
The design is deceptively compact. Externally, it presents as a modest and understated gym bag. But peek inside, and you’ll immediately see that this duffel, with its shocking 30-liter capacity, is Poppins-esque. There’s a dedicated shoe compartment on the side that accommodates up to a men’s size 14, though I prefer to use the bottom section for footwear to keep the main cavity flexible. There’s a slot for a 24-ounce water bottle, interior pockets for keys, AirPods, and other small essentials that tend to disappear into bag voids, and there’s still room for a change of clothes, a Theragun, and a dopp kit. Nothing about this bag feels over-engineered, but nothing feels missing, either.
Vietnam launches Hai Phong customs pilot from June 1
Is American men’s tennis in trouble as French Open begins?
Memorial Day Deals on Our Top Smart Bird Feeders, Camera Bags, and More
-
Tech1 week agoWhy Is Your Grill So Dumb? The Best Grills Set Temp Like an Oven
-
Tech1 week agoThis Solar-Powered Smart Sprinkler Keeps My Lawn Watered Without Any Power Cables
-
Fashion4 days agoICE cotton sees mixed trend as rain forecasts weaken sentiment
-
Fashion5 days agoUS cotton planting advances to 41%, concerns persist in Texas
-
Fashion1 week agoTurkiye’s current account deficit expected to widen in 2026: Minister
-
Fashion1 week agoIndia’s Pearl Global’s FY26 revenue crosses $521 mn milestone
-
Fashion6 days agoNigeria Kwara Garment Factory, KWS Garment Production Village ink pact
-
Entertainment1 week agoPrincess of Wales praised as ‘step ahead’ of royal family