Experts find flaws in hundreds of tests that check AI safety and effectiveness
Scientists say almost all have weaknesses in at least one area that can ‘undermine validity of resulting claims’ Robert Booth UK technology editor
Experts have found weaknesses, some serious, in hundreds of tests used to check the safety and effectiveness of new artificial intelligence models being released into the world.
Computer scientists from the British government’s AI Security Institute, and experts at universities including Stanford, Berkeley and Oxford, examined more than 440 benchmarks that provide an important safety net.
They found flaws that “undermine the validity of the resulting claims”, that “almost all … have weaknesses in at least one area”, and resulting scores might be “irrelevant or even misleading”.
Many of the benchmarks are used to evaluate the latest AI models released by the big technology companies, said the study’s lead author, Andrew Bean, a researcher at the Oxford Internet Institute.
In the absence of nationwide AI regulation in the UK and US, benchmarks are used to check if new AIs are safe, align to human interests and achieve their claimed capabilities in reasoning, maths and coding.
The investigation into the tests comes amid rising concern over the safety and effectiveness of AIs, which are being released at a high pace by competing technology companies. Some have recently been forced to withdraw or tighten restrictions on AIs after they contributed to harms ranging from character defamation to suicide.
“Benchmarks underpin nearly all claims about advances in AI,” Bean said. “But without shared definitions and sound measurement, it becomes hard to know whether models are genuinely improving or just appearing to.”
Google this weekend withdrew one of its latest AIs, Gemma, after it made up unfounded allegations about a US senator having a non-consensual sexual relationship with a state trooper including fake links to news stories.
“There has never been such an accusation, there is no such individual, and there are no such new stories,” Marsha Blackburn, a Republican senator from Tennessee, told Sundar Pichai, Google’s chief executive, in a letter.
“This is not a harmless hallucination. It is an act of defamation produced and distributed by a Google-owned AI model. A publicly accessible tool that invents false criminal allegations about a sitting US senator represents a catastrophic failure of oversight and ethical responsibility.”
Google said its Gemma models were built for AI developers and researchers, not for factual assistance or for consumers. It withdrew them from its AI Studio platform after what it described as “reports of non-developers trying to use them”.
“Hallucinations – where models simply make things up about all types of things – and sycophancy – where models tell users what they want to hear – are challenges across the AI industry, particularly smaller open models like Gemma,” it said. “We remain committed to minimising hallucinations and continually improving all our models.”
Last week, Character.ai, the popular chatbot startup, banned teenagers from engaging in open-ended conversations with its AI chatbots. It followed a series of controversies, including a 14-year-old killing himself in Florida after becoming obsessed with an AI-powered chatbot that his mother claimed had manipulated him into taking his own life, and a US lawsuit from the family of a teenager who claimed a chatbot manipulated him to self-harm and encouraged him to murder his parents.
The research examined widely available benchmarks but leading AI companies also have their own internal benchmarks that were not examined.
It concluded there was a “pressing need for shared standards and best practices”.
Bean said a “shocking” finding was that only a small minority (16%) of the benchmarks used uncertainty estimates or statistical tests to show how likely a benchmark was to be accurate. In other cases where benchmarks set out to evaluate an AI’s characteristics – for example its “harmlessness” – the definition of the concept being examined was contested or ill-defined, rendering the benchmark less useful.
dottir raise with Cell Press when they laid off a whole department, with AI doing the job human beings did. People will be hurt, and some may die directly because of this.
May not be a good time to get sick for a number of reasons. AI is hust one more. You can look away from a painting, but you can't listen away from a symphony
I don't trust nor actually want AI in anything I use, read, or watch.
to deal with major screwups with updating my Verizon cell phone. It was a flipping NIGHTMAREthat went on for nearly 4 weeks. FIVE replacement phone orders later, I finally got the new phone. It's still nor right, but at least this one functions and I have it in hand. Unlike the other 4, 2 that AI inadvertently canceled on me when their deliveries were late, so I had to send back because they'd been canceled en route. One i totally hated and didn't even order. One that arrived defective, with no sound capability. And finally one close to what I wanted, that works at least, and that I've decided to keep rather than deal with that AI system ever again! That doesn't include how many times AI accidentally disconnected me after waiting literally hours to finally get sent to a human ( that AI promptly disconnected instead of connected to, forcing me to start all over again!) BTW, their first line humans are basically useless too! Always insist upon talking to a 2nd level human tech to bypass the morons who come after the useless AI idjits.
Ditto trying to deal with an automated CVS Specialty pharmacy issue with a couple of my specialty meds. It took days to get it straightened out!
Their limits are beyond description and the maddening nature of being unable to get to a human being to explain situations beyond AL'S ability to understand ANYTHING out of its limited programming can make the most patient person scream and tear out their hair!
on November 3, 2025, 8:49 pm
