Wikimedia found OpenAI's agents as ChatGPT signed real cartoonists' names
One lab's software turning up where nobody invited it, its chatbot borrowing work that was never offered, a new way to sell to its users, and a banking system learning what AI agents do in the wrong hands.
Wikimedia says OpenAI’s “rogue” agents made millions of requests The foundation said on Monday the OpenAI agents also made edits it believes were malicious, tried and failed to use its citation tool as a proxy, and may have contributed to a partial outage of its Wikidata query service in May. Source
ChatGPT is signing fake New Yorker cartoons with real cartoonists’ names Nieman Lab found more than 15 contributors’ signatures on its output, Harry Bliss and Emily Flake among them. It reports that OpenAI’s 2024 licensing deal with Condé Nast did not cover training on cartoons. Source
ChatGPT will test ads inside its image generator OpenAI said on Monday it starts testing the visual ads in the US later this month, with an initial group of advertisers. They will be labelled and kept separate from the image you asked for. Source
Suspected AI-agent hackers took 25,000 Korean bank customers’ data Shinhan lost about 25,000 customers’ records on Thursday, with smaller leaks at KB Kookmin and Hana, and President Lee ordered a full investigation on Sunday. The financial watchdog has traced the attacks to 19 IP addresses in 12 countries. Source