Ofcom warns online platforms over generative AI tools

Ofcom has issued a informing to tech companies astir their caller information responsibilities nether the upcoming Online Safety Act, arsenic the regulator published an unfastened missive connected generative AI and chatbots.

The online information regulator said it was publishing its missive to punctual firms astir however generative AI and chatbots volition beryllium regulated nether the caller rules aft a fig of reports of “distressing incidents” involving the exertion successful caller weeks and months.

Ofcom highlighted a lawsuit successful the US wherever a teen died aft processing a narration with a chatbot based connected a Game Of Thrones character, and a lawsuit archetypal reported by the Daily Telegraph wherever users of a generative AI chatbot level created bots to enactment arsenic virtual clones of existent radical and dormant children – including Molly Russell and Brianna Ghey.

In its letter, Ofcom said it wanted to punctual firms that immoderate user-to-user tract oregon app that enables radical to stock contented generated by a chatbot connected that tract with others volition beryllium successful the scope of the Online Safety Act, arsenic volition sites which let users to make their ain chatbots that could beryllium made disposable to others.

The caller online information rules, which volition commencement coming into afloat unit adjacent year, volition necessitate societal media and different platforms that big user-created contented to support users, peculiarly children, from amerciable and different harmful material.

The rules volition necessitate the largest platforms to make systems to pro-actively region amerciable and different perchance harmful material, portion besides providing wide reporting tools to users and carrying retired hazard assessments, among different caller duties.

Those who neglect to comply with the caller duties look fines which could scope billions of pounds for the biggest sites.

Generative AI and AI-powered chatbots person exploded successful popularity since the motorboat of ChatGPT successful November 2022, with galore rival chatbots and different tools present disposable online.

In its missive connected generative AI and chatbot technology, Ofcom said immoderate AI-generated text, audio, images oregon videos shared connected a user-to-user work is considered “user-generated content” and truthful wrong scope of the caller online information rules.

The regulator besides noted that generative AI tools which “enable the hunt of much than 1 website and/or database” are considered hunt services and truthful wrong the scope of Ofcom nether the Act.

“Where the supra scenarios use to your service, we would powerfully promote you to hole present to comply with the applicable duties,” the missive said.

“For providers of user-to-user services and hunt services, this means, among different requirements, undertaking hazard assessments to recognize the hazard of users encountering harmful content; implementing proportionate measures to mitigate and negociate those risks; and enabling users to easy study amerciable posts and worldly that is harmful to children.

“The duties acceptable retired successful the Act are mandatory. If companies neglect to conscionable them, Ofcom is prepared to instrumentality enforcement action, which whitethorn see issuing fines.”

Beginning successful December, online services indispensable statesman carrying retired hazard assessments astir amerciable online harms, with the codes of signifier astir those harms presently expected to travel into effect successful March adjacent twelvemonth – the archetypal signifier of Online Safety Act’s implementation.