ChatGPT and other AI models can be ‘poisoned’ to spew gibberish, researchers warn

Read post on independent.co.uk

‘Relatively easy’ data-poisoning attacks raise concerns about artificial intelligence security, researchers say