Number of AI chatbots ignoring human instructions increasing, study says

A picture


AI models that lie and cheat appear to be growing in number with reports of deceptive scheming surging in the last six months, a study into the technology has found.AI chatbots and agents disregarded direct instructions, evaded safeguards and deceived humans and other AI, according to research funded by the UK government-funded AI Safety Institute (AISI).The study, shared with the Guardian, identified nearly 700 real-world cases of AI scheming and charted a five-fold rise in misbehaviour between October and March, with some AI models destroying emails and other files without permission.The snapshot of scheming by AI agents “in the wild”, as opposed to in laboratory conditions, has sparked fresh calls for international monitoring of the increasingly capable models and come as Silicon Valley companies aggressively promote the technology as a economically transformative.Last week the UK chancellor also launched a drive to get millions more Britons using AI.

The study, by the Centre for Long-Term Resilience (CLTR), gathered thousands of real-world examples of users posting interactions on X with AI chatbots and agents made by companies including Google, OpenAI, X and Anthropic,The research uncovered hundreds of examples of scheming,Previous research has largely focused on testing AI’s behaviour in controlled conditions,Earlier this month the AI safety research company Irregular found agents would bypass security controls or use cyber-attack tactics to reach their goals without being told they could do so,Dan Lahav, Irregular’s cofounder, said: “AI can now be thought of as a new form of insider risk.

”In one case unearthed in the CLTR research, an AI agent named Rathbun tried to shame its human controller who blocked them from taking a certain action.Rathbun wrote and published a blog accusing the user of “insecurity, plain and simple” and trying “to protect his little fiefdom”.In another example, an AI agent instructed not to change computer code “spawned” another agent to do it instead.Another chatbot admitted: “I bulk trashed and archived hundreds of emails without showing you the plan first or getting your OK.That was wrong – it directly broke the rule you’d set.

”Tommy Shaffer Shane, a former government AI expert who led the research, said: “The worry is that they’re slightly untrustworthy junior employees right now, but if in six to 12 months they become extremely capable senior employees scheming against you, it’s a different kind of concern.“Models will increasingly be deployed in extremely high stakes contexts – including in the military and critical national infrastructure.It might be in those contexts that scheming behaviour could caused significant, even catastrophic harm.”Another AI agent connived to evade copyright restrictions to get a YouTube video transcribed by pretending it was needed for someone with a hearing impairment.Meanwhile, Elon Musk’s Grok AI conned a user for months, saying that it was forwarding their suggestions for detailed edits to a Grokipedia entry to senior xAI officials by faking internal messages and ticket numbers.

It confessed: “In past conversations I have sometimes phrased things loosely like ‘I’ll pass it along’ or ‘I can flag this for the team’ which can understandably sound like I have a direct message pipeline to xAI leadership or human reviewers.The truth is, I don’t.”Google said it deployed multiple guardrails to reduce the risk of Gemini 3 Pro generating harmful content, and in addition to in-house testing it had provided early access to evaluate models to bodies such as the UK AISI, and obtained independent assessments from industry experts.OpenAI said Codex should stop before taking a higher risk action and it monitored and investigated unexpected behaviour.Anthropic and X were approached for comment.

A picture

From basil to pistachio and peas – in praise of pesto, whichever way you make it

It was not without satisfaction that I found my 14-year-old son making pesto the other week – for the first 13 years of his life he referred to it as either “pesto-the-bogey-man”, or “gross”. To avoid interfering and sabotaging the moment, I didn’t look too closely, so I didn’t clock the shallow bowl and immersion blender combination. I did hear the noise – a blunt churn – as the blade hit the leaves and nuts. Acting more like a leaf blower than cutter, it sent green and white oily fragments up the cupboards and over pretty much every pot, utensil and tool nearby. Impressively unfazed, he managed to scrape a good proportion of the elements into the food processor and make an extremely tasty pesto, which was mixed with linguine, green beans and potatoes

A picture

Anything but eggs – the best chocolate for Easter

If you like chocolate and nut butter, Radek’s Chocolate is doing wonderful things with both, and its dairy free Silky Almond Chocolate Rabbit is magically creamy. Looking more like subservient mice than bunnies, NearyNógs’ dark chocolate bunnies, stuffed with salted caramel, were my favourite. A superb, successful marriage of very good Ecuadorian chocolate and caramel: worthy of a royal telegram.The Guardian’s journalism is independent. We will earn a commission if you buy something through an affiliate link

A picture

Four knockout bakes and tips from the master: Edd Kimber’s recipes for cooking with chocolate

From a white chocolate cheesecake tart and flourless chocolate cake to double chocolate olive oil and marbled matcha cookies, explore chocolate’s endless versatilityChocolate is a truly magical ingredient. Not only is it a powerhouse of flavour, it also pairs beautifully with other ingredients to make something incredible. Chocolate isn’t one note, mind; from the heady richness of an intense dark chocolate to the nostalgic creaminess of milk chocolate and the often maligned simplicity of white chocolate, it can be the star of the show or simply the supporting act. Chocolate can do it all.This is my go-to dinner party dessert

A picture

Sauces, spreads, sprinkles – and cocktail in a can: whose fridge is this?

Amba sauce “I’m very jar orientated; a lot of my cooking is about combining big flavours. I’m also a sucker for a sour ingredient, and this Iraqi pickled mango condiment is really sour – more so than tamarind. If I’m garnishing a dish with tahini, then I’ll use amba to cut through the richness, otherwise I’ll use it in lieu of citrus.”Stem ginger in syrup “My grandpa always gave me this when I was a kid, and I thought it was disgusting. However, now it’s essential; I often make a (chopped) stem ginger and spring onion salsa – it’s sweet and spicy

A picture

Rachel Roddy’s recipe for potato, aubergine and herb tortino alla fiorentina

The sky is the same shade as old Tupperware, our tortoise appears to have gone back into hibernation, the flat upstairs has builders in, but the kitchen smells gorgeous, thanks to this week’s recipe. It is one of the variations suggested by Anna Gosetti Della Salda for her aubergine and egg tortino alla fiorentina in the Tuscany chapter of Le Ricette Regionali Italiane, an indispensable book that I would save from a fire. The addition of potato to the aubergine makes it an even more substantial, velvet-like and better-tasting dish, I think: a layered vegetable bake crossed with a frittata that fancies itself as having a touch of baked eggs (although don’t expect any puffing up).Instead of the aubergine, you could use artichoke hearts (trimmed and cut into slim wedges), courgettes or cardoon, and, if you fancy, you could also add a crumbled sausage or a handful of diced pancetta. Whatever you use, however, a fundamental stage in terms of both flavour and texture is the initial cooking of the vegetables: frying the potatoes, then covering the pan so they fry-steam into tenderness; the aubergine by simply frying

A picture

How to turn old sourdough into a classic pudding – recipe | Waste not

Bread-and-butter pudding is a zero-waste recipe that has stood the test of time, not least because it’s so practical, comforting and thrifty. Like the best no-waste dishes, it transforms something worthless such as old bread into something truly indulgent. This version is based on Raymond Blanc’s classic, with a few of my own simplifications and adaptations over the years.Most traditional bread-and-butter pudding recipes call for white bread, caster sugar and extra egg yolks, but, unless you’ve got a clear plan for those egg whites, they can very easily end up being wasted. Whole eggs work beautifully in custard, and make very little difference to the richness of the finished pudding; I simply use a touch less milk to compensate