Can You Sue Over Assurances Made by Company's Customer Service AI Chatbot?

Maybe, but not in this particular case, a federal court rules.

From Magistrate Judge Alex Tse's opinion today in Smith v. Substack, Inc. (N.D. Cal.):

Prior to the filing of this case, there was a series of interactions between Smith and the Doe defendant. Smith alleges that this unknown defendant posted unflattering statements about Smith on Cancel Watch, a blog site hosted by Substack. Smith initiated contact with Substack, including twenty to thirty "complaints and queries by email" between July and September 2023, all to no avail. Substack did not respond to any of Smith's emails regarding Cancel Watch.

In February, March, April, and May of 2024, Smith asked a series of questions to a chatbot found in the support section of Substack's website. Smith asked, "do you respond to complaints," and the chatbot responded, "Yes, we respond to all complaints." He also asked, "do you respond to every complaint," and "do you always [all of the time], respond to complaints?," to which the chatbot responded with the same answer or a very similar one. Id. Smith then asked, "does Substack respond to emails?" and "Will you certainly respond to emails?," the chatbot said, "Yes, Substack responds to emails" in response to both inquiries. Smith alleges that the answers from the chatbot are the same for "queries," and that Substack says it will respond to reports. However, regardless of its chatbot's replies, Substack itself never did respond to Smith's inquiries, or to his follow-up inquiries asking why the company was ignoring him….

Smith sued Substack under a promissory estoppel theory, which is related to breach of contract. No, said the court:

The chatbot's responses to Smith, however, are not sufficiently definite to give rise to a claim of promissory estoppel. The chatbot said that Substack would respond to complaints, emails, and queries. However, the chatbot did not say anything about how Substack would respond, or when. Without those essential terms, the Court cannot discern when Substack is in breach of its obligations. See White v. J.P. Morgan Chase, Inc. (E.D. Cal. 2016), aff'd (9th Cir. 2017) (finding that plaintiffs failed to allege promissory estoppel when the promises at issue "were fatally uncertain because they contained no essential terms"). Theoretically, Substack is still not in breach given that no timeframe for a response was promised….

Second, "detrimental reliance is an essential feature of promissory estoppel." Reliance might be found when a "promisee suffered actual detriment in foregoing an act, … or in expending definite and substantial effort or money in reliance on a promise." … Smith contends that he relied on the promises made by the chatbot because he had exhausted all other methods seeking a response from Substack. Smith states that his reliance "was under Substack's assistance or help" because the chatbot features in the support section of their website. Substack argues Smith has failed to allege detrimental reliance, as the [Complaint] does not state any facts showing that Smith changed his position or acted to his detriment in response on a promise from Substack.

In his opposition, Smith alleges that he filed a new criminal complaint on April 23, 2024, after receiving assurances from the chatbot. Smith would not have filed the new criminal complaint but for the chatbot's assurances that Substack would respond to emails and complaints. Several weeks later, after not receiving a reply, Smith "closed his complaint." Smith alleges that his time and police time were wasted, and he experienced emotional distress as a result. Additionally, he alleges nominal damages to recover for wasted expenses in connection with a phone call to the police.

In deciding a Rule 12(b)(6) motion, a court is limited to the complaint. New facts raised only in the opposition may be considered for the purposes of deciding leave to amend, but not the Rule 12(b)(6) motion. As such, in deciding the motion to dismiss, the Court will only consider the allegations in the [Complaint].

Smith's [Complaint] here fails to allege facts to show detrimental reliance. ⁵ Plaintiff merely contends that he relied on the promises of the chatbot and explains why he relied. But Smith does not allege what actions he took or did not take, or what efforts or money he expended in reliance. Without that, the [Complaint] does not contain any facts to show any actual detriment to Smith as a result of his reliance on the communications from the chatbot….

And the court declined to let Smith amend his complaint:

Reliance can be found when a party "expend[ed] definite and substantial effort or money in reliance on a promise." Smith alleges in his opposition only that he filed a new police report and then dismissed it, which wasted his time and led to emotional distress and a wasted phone call. These new allegations do not rise to the level of substantial effort or money expended in reliance on the chatbot's responses and were made in response to Substack's assertion that Smith failed to show reliance. Because these are the best facts (if taken true) in support of reliance that Smith can allege, another amendment would unlikely save the claim.

Moreover, Smith filed the new criminal complaint on April 23, 2024, but alleges reliance on chatbot responses in February, March, April, and May. To the extent that Smith claims to have relied on responses which came after he filed the new criminal complaint, Smith has still failed to allege any reliance at all, substantial or not. Thus, the Court finds that allowing another amendment would be futile.

Smith has already had three opportunities to plead these claims. In light of the previous amendments, the Court has broad discretion to deny leave to amend and does so. Smith's claim for promissory estoppel is dismissed with prejudice….

Benjamin D. Margo and Maura Lea Rees (Wilson Sonsini Goodrich & Rosati) represent Substack.