Thanks to the European Union's new AI Act and other legislation winding its way through the U.S. courts, AI is about to get regulated, hard.

"Just So You Know, I'm an AI"

The AI Act restricts how AI are used that potentially threatens people's fundamental rights, such as healthcare, education, and policing. Of particular note is a focus on “subliminal, manipulative, or deceptive techniques to distort behavior and impair informed decision-making."

On top of all that, AI will be required to clearly identify itself as an AI, a potential problem for persona- and companion-style AI that act as people in their interactions. They may soon become less personal as new guardrails are implemented so that they constantly remind users that they are AI.

Notably, the act also bans AI systems that infer characteristics, like political opinions or sexual orientation. This is a big deal with generative software that writers letters -- AI won't be allowed to guess a person's gender anymore from their name alone, so no more using AI to write form letters to large groups of people.

"Here's My Homework"

Not only will AI have to identify itself as such, it will need to be more transparent about how it came to generate content , including a digital "paper trail" about its training data. This can potentially affect everything from how GitHub Copilot scoops up coding data (Doe et al. v. GitHub, Inc. et al. ) to the text it ingests (Silverman et al., v. OpenAI Inc et al. , yes that Sarah Silverman ) to the art styles it mimics (Andersen v. Stability AI et al. )

"Please Speak to My Manager"

Thanks to the AI Act, there will be a recourse in the European Union for those who feel AI has harmed them. Data privacy is of particular concern, especially when the AI may be at risk of a breach, or simply uses the information it has about individuals inappropriately. To see how much has changed, only a year ago I asked ChatGPT to write a bio and it got some things distressingly wrong (conflating controversial topics I reported on with me being involved in said controversial topics). Now it refuses to write a bio for anyone. Gemini and Copilot have followed suit.

These changes will likely reduce the "genie-like" effects of generative AI where you could ask them anything and they'd do their best to answer. Instead, we'll get significantly more focused (and likely, less useful but more reliable) AI like Copilot, who transitioned from ChatGPT-4 to ChatGPT-Turbo, a much less friendlier bot . Add all this up, and it looks like the Wild West days of AI companies doing whatever they want is rapidly coming to a close. And that's a good thing.

Please Note: The views and opinions expressed here are solely my own and do not necessarily represent those of my employer or any other organization.