Krista L. Baughman
Partner
Baughman Law
1255 Treat Blvd., Suite 300
Walnut Creek , CA
Phone: (925) 310-6033
Email: krista@baughmanlaw.com
Brooklyn Law School; Brooklyn NY
Krista's practice focuses on First Amendment and anti-SLAPP law and litigation.
What qualifies as "speech" in the age of AI and social media? Are algorithmically generated feeds deserving of First Amendment protection?
These are the tough questions at the heart of Meta Platforms, Inc. v. Bonta, a case challenging California's "Protecting Our Kids from Social Media Addiction Act" (SB 976), in which Meta, TikTok and Google seek to prohibit Attorney General Rob Bonta from enforcing sections of the act which would bar them from sending minor users "addictive feeds" that are based on the minor's prior interactions with their platforms. This month, U.S. District Judge Edward Davila rejected the platforms' request for a preliminary injunction, holding that the platforms are unlikely to show that the personalized feed provisions infringe their First Amendment rights.
The court's reasoning raises a fundamental--and philosophical--question about the meaning of "speech" in today's digital age: Are all algorithmically generated feeds necessarily "expressive"? The answer will determine whether platforms will have First Amendment rights in the feeds they publish to users.
The Supreme Court has provided some guidance in this area. In Moody v. NetChoice LLC, 603 U.S. 707 (2024), the Court discussed how social media companies, acting through human employees, enact community guidelines or community standards to define what kind of content is--or is not--acceptable on their platform. For example, many platforms prohibit the publication of violent content, content involving sexual exploitation, or content they deem to be "hate speech." The Court reasoned that when platforms use their community guidelines to compile user feeds, they make "expressive choices," akin to the editorial discretion employed by newspapers or parade organizers in determining what speech will be included or excluded from a final product. This type of curation, the Court held, is speech protected by the First Amendment from government interference.
But Moody's reasoning does not necessarily answer the question of whether personalized or "addictive" feeds are similarly expressive. Indeed, Moody included a carve out that anticipated this issue: "[w]e therefore do not deal here with feeds whose algorithms respond solely to how users act online--giving them the content they appear to want, without any regard to independent content standards." See also Moody (Barrett, J., concurring) ("But what if a platform's algorithm just presents automatically to each user whatever the algorithm thinks the user will like--e.g., content similar to posts with which the user previously engaged?") This skepticism suggests that the Supreme Court may not extend First Amendment protection to personalized feeds.
Which is precisely what Judge Davila found in Meta Platforms v. Bonta--at least, on a preliminary basis. Judge Davila noted that in generating personalized feeds, platforms "rely on predictive algorithms that incorporate users' past watch history and other data and then suggest content that the algorithms anticipate will be engaging, or 'interesting,' to users." According to the court, "[t]his decision is not an 'expressive' message; it is merely a mirror that reflects back to users their own perceived interests." Since personalized feed algorithms make user-interest determinations "divorced from any understanding of the underlying content," they are essentially "'dumb pipes' that merely emit what they are fed" and are therefore not deserving of First Amendment protection.
The court considered the platforms' argument that personalized feeds are speech because human employees make various expressive decisions while developing and monitoring the recommender systems. The court was not persuaded, reasoning that "the decision to 'convey to users ____ because we think whatever fills in that blank is interesting, relevant or trustworthy'" is not itself an expressive choice. The court contrasted this with the content moderation decisions discussed in Moody, which effectively fill in the blank with the platform's own policy positions (e.g. "we don't want nudity on our platforms"), and thus "carry with them a moral valence" that rises to the level of expression. On this basis, the court denied the platforms' preliminary injunction request.
Will the ruling hold up on appeal? The debate is vigorous. One side of the discussion, as seen in Davila's reasoning, is grounded in first principles: after reciting that "[t]he touchstone of First Amendment speech rights is [] the protection of expression," Davila questions whether reliance on "algorithms' number-crunching capabilities" to predict user preferences serves that noble end and is thus "expression" at all. This reasoning draws from the truth-seeking, self-governing rationale for the First Amendment--the Madisonian concept that free expression is essential to democratic self-rule and critical thinking--and displays a skepticism that a commercial product designed to maximize user engagement is worthy of being called speech.
Others contend that the choice to use a delivery mechanism in the first place is an expressive, editorial decision that invokes the First Amendment, regardless of the precise nature of the algorithm. This argument contends that if courts were to hold otherwise, they would be putting their thumb on the scale in a subjective, viewpoint-based way, and stresses that it is problematic to entrust in government the power to decide what content delivery decisions are "expressive enough" to invoke the First Amendment's umbrella.
In her Moody concurrence, Justice Barrett poses hypotheticals that further complicate the analysis: What if "a platform's owners hand the reins to an AI tool and ask it simply to remove 'hateful' content?" In that case, has a human being with First Amendment rights made an inherently "expressive" choice? Barrett notes that "technology may attenuate the connection between content-moderation actions (e.g., removing posts) and human beings' constitutionally protected right to "decide for [themselves] the ideas and beliefs deserving of expression, consideration, and adherence."
Ultimately, there is no easy answer to these questions. The rights bestowed by the First Amendment are not God-given; they were contemplated, negotiated and determined by human beings at a time when algorithmic speech was unforeseeable. Whether and to what extent these rights will be expanded to cover increasing amounts of computer-generated outputs will be a defining question for the Supreme Court and legislating bodies--and for We the People.