Meta’s new AI privacy policy allows targeted ads, possibly political ads, based on chats with bots

Some Instagram users may start seeing that political advertisements mirror what they say and search on the social media platform.

The change is part of Instagram’s newest privacy policy that allows the company to use conversations with Meta’s artificial intelligence to personalize experiences and ads.

Interactions on social media platforms have long dictated each user’s personalized algorithms. The new Instagram policy update adds AI interactions to the mix.

Instagram, owned by technology giant Meta, rolled out its new AI initiatives on Dec. 16. If users interact with Meta AI, which includes Instagram’s search bar, the social media platform will use that interaction to tailor ads and content relevant to the user.

Meta’s new policy drew opposition from an array of advocacy groups.

A coalition of 36 privacy, consumer protection, children’s rights and civil rights advocates and researchers urged the Federal Trade Commission to stop Meta’s plan.

“The danger of regulatory inaction at this critical moment, when generative AI is driving an unprecedented expansion of commercial surveillance, could be catastrophic and irreversible,” the coalition said in a letter to the FCC.

When rolling out the new policy, Meta used an interest in hiking as an example of how AI interactions could turn into targeted content.

If a user chats with Meta AI about hiking, the company said, its algorithm learns that the user has an interest in hiking. The user may then see recommendations for hiking groups, posts about trails or ads for hiking boots.

Meta said that this is not significantly different from someone posting or linking similar content.

This AI advancement could also have political implications. Meta said that if a user has a conversation with Meta AI about political views, among other topics ranging from sexual orientation to trade union membership, Meta will not use this to influence ads. This is because they “could have special protections under the laws of your jurisdiction.”

If a user has a direct conversation with Meta AI about politics, that will not be used. But it may indicate other things, as there are plenty of ways companies can infer a user’s political views, said Darrell West, director of the Center for Technology Innovation at the Brookings Institution.

If an Instagram user talks to Meta AI about an interest in renewable energy, for example, Meta could use that to tailor related content and ads. Because this conversation is outside the political views prohibition, the information could be used to target political content.

The same would be true if the user makes posts or comments about renewable energy.

However, if the user explicitly voices support for a specific political candidate because of their views on renewable energy, that conversation would be off-limits because of its political nature.

Scott Babwah Brennen, director of the Center on Technology Policy at NYU, said it’s not so terrible that he could see ads for a candidate that represents his interests.

While he said he is always a little concerned about how personal data is used in advertising and beyond, he is sympathetic to the argument that there might not be a significant harm if it delivers relevant ads to consumers.

But using conversations with AI to influence ads will “allow targeting at such a precise level,” Mr. West said, almost becoming individualized to the point that taking detailed information about users is normalized.

A high degree of targeting could also exacerbate political polarization and extremism in the U.S. by inundating people with specialized content rather than other viewpoints.

“You can really play to the crowd and tell them exactly what you know they want to hear, so it’s a very personalized type of politics, which could lead to even greater extremism than what we already see,” Mr. West said.

People imagine that their conversations with AI are private and confidential, Mr. West said, adding that they will discover that is not the case.

Few national laws are in place to oversee the collection of data through AI conversations and using it to influence political ads. Some states, such as Minnesota, have taken initiatives to counter this.

The Minnesota Consumer Data Privacy Act grants consumers the right to opt out of businesses selling, profiling and using targeted advertising with personal data. It also gives consumers the right to question automated decisions made about them, including those facilitated by AI.

President Trump recently signed an executive order that bars states from regulating AI, setting the stage for court battles over regulating the new technology.