We need AI regulation, but we also need to get it right

New York's FAIR news act raises a LOT of questions

The New York State Legislature sent a bill to the Governor’s desk that acts as a clunky and bad example of what should be a good thing: AI regulation.

The New York fundamental artificial intelligence requirements in (FAIR) news act requires any news organization to clearly state when a piece of news is “substantially composed, authored, or otherwise generated” through Gen AI. It does offer an exemption to work that is eligible for copyright protection, so it’s unclear how much actual material would need to be labeled. It also comes just as another AI disclosure law focused on “synthetic performers” in advertising content went into effect.

But the single line in this bill that gives it teeth raises a number of questions:

• What does it mean to be “substantially composed”?

• Does using a tool like Grammarly alter the content in such a way that makes it suspect?

• Even if we can figure out a definition of “substantially composed”, how will the enforcement body even know if something has been “substantially” generated by AI? The detection tools are notoriously suspect.

• What does this mean for other tools like AI search or data analysis that reporters often use earlier in the reporting process?

• What defines disclosure? If a TV news show adds a general line to the end of the show saying something like “portions of this broadcast were written with the aid of AI,” would that be enough? And if so, is that truly a protection for the audience, or just more noise?

The bottom line here is that this is a bad law and it’s aimed at the wrong target. The intent of the law is sound, as the sponsors note how AI can contain false content and “plagiarizes by deriving its content from original source material without permission or proper citation.”

But false content is an issue with editing and fact-checking, not with AI itself. A news organization should be living and dying on its reputation, which means making sure its reporting is accurate. Relying on a bad reporter or AI results in the same bad information that could kill an organization’s reputation.

If the legislature is concerned with how AI acquires its content, then the regulation should be at the training level, why hold the news industry accountable for how the LLMs train their system?

Also, as any PR pro will tell you, news organizations are now only one part of the overall news ecosystem, with people getting information from influencers and other accidental journalists. At the AI x Journalism conference, I met the founders of the Independent Journalism Atlas, which is tracking journalists across all platforms. These aren’t always traditional news organizations like the New York Times and ABC News, but rather people with Substacks like this one or Twitch feeds.

We do, in fact, need AI regulation, and it should be aimed at the larger players, not at the users. In the Pope’s encyclical on AI, he points out (paragraph 95):

In many cases within the digital context, control over platforms, infrastructure, data, and computing power does not rest with States, but with major economic and technological actors. These entities effectively set the conditions for access, determine the rules of visibility and shape the very possibilities for participation. When such power is concentrated in the hands of a few, it tends to become opaque and evade public oversight, increasing the risk of distorted forms of development that give rise to new dependencies, exclusions, manipulations and inequalities.

Over in the New York Times, John O’Farrell, former General Partner at the Silicon Valley Powerhouse Andreesen Horowitz, echoes this sentiment saying that as the industry is now “could concentrate economic power in ways that would make the Gilded Age look quaint.” He says that the power players who are fighting AI regulation are making a dangerous mistake, we do need regulation. But it needs to be done correctly.

Bad regulation could hobble a transformative technology. Our politicians have not always distinguished themselves, with some unsophisticated, ill-informed attempts at technology regulation. But you don’t achieve balanced, intelligent regulation by silencing debate — you get it by engaging seriously and earning trust.

We do need serious regulation of the AI industry, but we also need to do it thoughtfully and target the right industries. I do not think the FAIR Act reaches that standard.