Lessons from the Associated Press AI guidelines

The Associated Press has published guidance on how it will deploy generative artificial intelligence in its news operation and stylebook, with useful applications beyond the business of journalism.

Why AI guidelines matter

In the newsroom, AI throws up some incredibly knotty topics ranging from the veracity of the facts through to the livelihood of writers (among other creative or content-led professions). For now, the AP is – like many other companies including, most vocally, Coca-Cola – in an experimentation phase, the service explains.

In the first instance, the existence of rules at all is a good first step. Formal regulation will take time to put in place and enforce, so companies that are dedicating serious time and thought to their use of the technology will be better placed should rules come into play.

“Answering these questions at the outset,” the AP states, “will save your newsroom a lot of time and energy compared to retroactively figuring out responsible use of a tool after purchasing it, training it, and using it.”

What to know

• The main thrust of the ideas surrounds editing, with the AP saying “material produced by artificial intelligence should be vetted carefully, just like material from any other news source.”
• The AP has a licensing agreement with Open AI, which is known for its ChatGPT system, but staff are not allowed to “use it to create publishable content.” 
• Elsewhere, AP's standards stipulate that no element of photo, video, or audio can be edited (either by adding or subtracting) using AI.
• In a nod to the ability for bad actors to promulgate AI-generated fakes or “deep fakes”, the AP will not publish the image unless it is the focus of the story and it will then be clearly labelled as AI-generated.

Key quote

“You still have to be a journalist” – Amanda Barrett, AP vice president for standards and inclusion.

Sourced from AP, Poynter, WARC