The accidental creation of an AI system capable of creating sophisticated fake news stories will be withheld from the open source community over fears it will be abused for malicious purposes.
Researchers at OpenAI institute said they were attempting to create an algorithm that could produce natural sounding text based upon extensive research and language processing, but soon realised it was capable of creating fake news stories taking cues from the 8 million web pages it trawled to learn about language.
Given how convincing some of these stories are, it's believed the system could be exploited to spread malicious or inflammatory content.
"We started testing it, and quickly discovered it's possible to generate malicious-esque content quite easily," said Jack Clark, policy director at OpenAI, speaking to the BBC.
OpenAI's researchers used content posted to link sharing site Reddit that had achieved a "karma" score of 3 or more to make the sources more reliable. It then uses these sources to write stories, making up attributions and quotes to make them sound more convincing.
However, researchers found that the occasionally included inaccuracies, with names and places being used incorrectly. The BBC references one story where a person named "Paddy Power" led a protest, for example.
The research will now be used as a platform to demonstrate that AI applications should be used carefully and to launch a debate about whether it should ever be used for things like news writing.
"It's not a matter of whether nefarious actors will utilise AI to create convincing fake news articles and deepfakes, they will," Brandie Nonnecke, director of Berkeley's CITRIS Policy Lab told the BBC.
"Platforms must recognise their role in mitigating its reach and impact. The era of platforms claiming immunity from liability over the distribution of content is over. Platforms must engage in evaluations of how their systems will be manipulated and build in transparent and accountable mechanisms for identifying and mitigating the spread of maliciously fake content."
