My Ai Did It: Intermediary Liability for Ai?

• Publication year: 2022
• Author: Ryan E. Long

MY AI DID IT: INTERMEDIARY LIABILITY FOR AI?

Ryan E. Long
Long & Associates

More companies are outsourcing their decisions to employ artificial intelligence ("AI"). For example, the market for AI powered facial recognition technology was valued at $3.72 billion in 2020 and is projected to grow to $11.62 billion by 2026.¹ Similarly, the marketplace for the use of AI in assisting consumer spending decisions is slated to grow at a Compound Annual Growth Rate of 34.4% to $19.9 billion by 2027.²

This has raised the question in recent litigation of whether "my AI did it" can be an acceptable defense to claims including, among others, privacy violations or even copyright infringement. In some cases, vicarious or secondary liability doesn't apply. This article addresses some more recent case law on the issue involving AI. It also summarizes the Algorithmic Accountability Act of 2022 and its potential effect on liability for reliance on AI-based decision making.

VICARIOUS LIABILITY

A defendant can be vicariously liable for copyright infringement by others under certain circumstances. "[C]ontributory infringement is based upon 'the common-law doctrine that one who knowingly participates or furthers a tortious act is jointly and severally liable with the prime tortfeasor."³ Thus, a person—even a company—can commit "contributory copyright infringement by 'intentionally inducing or encouraging direct infringement[.]'"⁴ In all cases, there must be a predicate act of "direct copyright infringement," or else "secondary liability cannot be maintained."⁵

Similarly, "[a] corporate officer can be held liable under a theory of vicarious copyright infringement where he profits from direct infringement while declining to exercise a right to stop or limit it."⁶ A requirement in this analysis is that the officer is a "moving, active, conscious force behind the defendant's infringement."⁷ Knowledge—either actual or constructive—of the infringement is required. In other words, there is no "ostrich in the sand" defense to vicarious liability for copyright infringement.

Immunity under the Communications Decency Act (a/k/a "Section 230") and the safe harbor provision of the Digital Millennium Copyright Act ("DMCA") are based on the foregoing principles. Under Section

[Page 50]

230, an interactive computer service provider like Twitter is not vicariously liable for defamatory or other potentially unlawful content—save for some statutory exceptions—posted by others so long as the provider doesn't modify the "basic form and message."⁸ Once Twitter becomes a "conscious force" in modifying the content of the message, as opposed to merely keeping it up or taking it down, liability under Section 230 can exist. Under the DMCA, "Internet service providers [are exempt] from liability for copyright infringement if the provider is unaware of the infringement and acts expeditiously to remove the copyrighted material upon notice."⁹ Unlike Section 230, where awareness doesn't confer liability, an internet service provider under the DMCA can be liable for infringing content if they are conscious of it but do nothing.

As we will see below, some federal statutes not only require knowledge or intent for liability to exist—but also lack of remedial conduct, similar to the DMCA. The question arises whether a company's use of AI can change this legal landscape so as to avoid liability.

OUTSOURCING USING AI

Various recent decisions directly or indirectly touch upon this issue. In Lopez v. Apple, for example, a class action was filed against Apple in connection with its pre-installed software called "Siri," which is a voice activated "intelligent assistant."¹⁰ This artificial intelligence-based software is supposed to only listen to, record, and share user conversations when they give consent—by saying a "hot word,"...