Even tech bros are warning that consensus needs to be found on AI safety before it spirals out of control - Jayne Dowle

Artificial intelligence agents going ‘rogue’, taking actions outside their human-led instructions, is the terrifying stuff of nightmares. I’m trying not to be alarmist, but we should prepare ourselves for this scenario imminently, technology experts warn.

A network of such rogue AI agents could conceivably seize control of the internet within six months to a year if companies do not focus more on implementing safeguards, cautions Dario Amodei, CEO of Anthropic, the San Francisco-based firm behind the Claude AI model.

Blame them if you like, but when even those who are benefiting the most financially from this rapid development of technology are saying artificial intelligence must slow down fast, we should all sit up and listen.

Claude offers four levels of AI, from basic chatbots and writing and administration tasks to what it describes as “demanding, long-horizon agentic workflows that require minimal step-by-step human guidance”. Its overseer is urging restraint, so where does that leave us?

I may be a dinosaur, but after all, my professional life as a writer and journalist is at stake as the AI revolution threatens to make my living redundant, but I refuse to be complicit. I’m going down fighting.

Accepting that even phoning our local GP surgery now involves conversing with a robot named Emma, I know I cannot avoid AI’s encroachment into everyday life.

However, by no means am I engaging with it by my own volition. I actually cringe silently - and not always silently, I’ll admit - when I hear someone say they’ve put a basic question, such as recommendations for restaurants, or the average call-out cost for a local plumber, ‘into Chat GPT’ without attempting to figure it out themselves, perhaps assisted by a few judicious humanly-generated independent internet searches.

This might sound harsh, but if we’re already allowing our brains to turn to mush and let machines do our thinking we deserve what’s coming to us. This may well be a pointless fight against an overwhelming force, but I’m making a personal stand and holding out until the last.

Meanwhile, we need our politicians to take the lead, quickly. Ahead of them is a global task so huge it cannot even be comprehended by human intelligence. But that is no excuse. This is no time to shy away from a rapidly-evolving situation, cowering in fear.

A cross-party group of MPs and peers is already warning that an AI Bill is needed “to address the scale and seriousness of threats posed to human rights”.

Meanwhile, the Joint Committee on Human Rights (JCHR) listed in a report a range of human rights risks which existing regulatory and legal frameworks “are ill-equipped to prevent or respond to”.

We are not talking here about nebulous notions of freedom and liberty, but our very existence as human beings; the way we live, control our decisions, plan for the future and protect ourselves. The stuff of life, in other words.

Whilst our own prime minister is just one world leader wondering what on earth to do, serious times demand a serious leader. Now is not the time for diversions or populist stunts. He needs to put himself - and his government - on a war footing, and rather than point-score with opposing political parties, hammer out an agreed UK approach.

All 195 countries in the world - from the mighty USA, Russia and China, to the Vatican, the smallest nation state - will come at AI from differing angles, dependent on a huge range of factors, including existing take-up of technology and economic health to whether they are currently fighting a war or not or generally believe in a higher holy being.

This is where human frailty, in the form of rampant egos and self-protection threaten to become the undoing of any accord. Would you want to sit down President Trump, Vladmir Putin and Xi Jinping even in a virtual room together and try and get them on the same page?

However, as the global consensus on safety standards in aviation proves, it is possible for world leaders to agree on what is important for human safety. Remember, even the tech bros who developed AI are now warning it is becoming out of control. If this isn’t a cry for help, I don’t know what is.