Facebook Doesn’t Have a Duty to Prevent a Murder–Godwin v. Facebook

In 2017, Steve Stephens murdered Robert Godwin Sr. On the day of the murder, Stephens made the following post to Facebook:

FB my life for the pass year has really been fuck up!!! lost everything ever had due to gambling at the Cleveland Jack casino and Erie casino…I not going to go into details but I’m at my breaking point I’m really on some murder shit…FB you have 4 minutes to tell me why I shouldn’t be on deathrow!!!! dead serious #teamdeathrow.

(Stephens also posted a video of the murder to Facebook, but the court doesn’t directly address the legal implications of that fact).

The estate sued Facebook for Stephens’ murder of Godwin. The court summarized the estate’s legal theories:

(1) the common law negligence claim, based on the failure to warn Robert Godwin of Stephens’s dangerous propensity of which Facebook was aware through its data-mining practices, which is the underlying negligence theory upon which the wrongful death and survivorship claims arise and (2) the civil-right-of-recovery claim based on R.C. 2921.22 and 2307.60 stemming from Stephens’s message posted to his social network page “minutes” before Robert Godwin’s tragic and senseless murder.

The trial court granted Facebook’s motion to dismiss. The appeals court affirms, with a troubling concurrence.

The Court Opinion

The central legal question is whether Facebook owed a duty to Godwin. In the context of standard commercial businesses like Facebook, the duty only applies to business who have a “special relationship” with the plaintiff–businesses do not owe abstract duties to everyone in the world.

The court says the complaint didn’t properly allege a special relationship between Facebook and Godwin. The estate’s basic argument is that Facebook’s data-mining practices should have informed Facebook of Stephens’ murderous intent, but the court (using a highly technical analysis of tort doctrine) says that Facebook never “takes charge” of its users’ conduct enough to assume the legal duty to control their behavior. This allows the court to distinguish the limited circumstances where vendors may in fact have legal duties to prevent their customers from committing harm to third parties, such as doctors or psychiatrists. The court also denies the claim that Facebook had a duty to report Stephens for making a terroristic threat because Stephens’ statement that he may “do some murder shit” wasn’t a credible threat. The court expressly declined to apply Section 230 to this case.

A concurring judge thinks that Facebook had duties here because “negligence is in the air.” Thus:

As a matter of policy, public safety should be of primary concern, which is why we have tort law. I truly do not see Facebook’s issue. It had information of a potential crime. By acting it might have saved a life. Of course, we will never know, but that is why we give individuals their day in court… Only when legal and moral duty diverge can courts hear a call for movement and reform.


The Fact Problems In This Case. This case involves a terrible tragedy, and I don’t want to detract from the pain and grief the family is feeling. However, focusing purely on the legal questions, this case has some obvious problems:

  • Stephens’ Facebook post was incoherent and, at most, vaguely anticipatory of future crimes. It doesn’t actually say that he is imminently going to commit any crimes, and there are likely millions of similar posts with the same speculative gestures towards future criminal activity that never result any crimes actually being committed. There’s no way that even the best “data-mining” technology would have accurately identified this post as worrisome (or, if it did, it would also likely create millions of false positives). For more on the false positive problem, see this TSF case study on Twitter and self-harm.
  • The complaint indicates that Facebook should “know” about this post because it does data-mining. But data-mining to anticipate who will commit future crimes is straight out of Minority Report. Asking Facebook to predict what Stephens would do is exactly what the movie cautioned us not to do.
  • Assuming that Facebook “knew” that Stephens planned to commit murder, who should Facebook notify? It would have been impossible for Facebook to warn Godwin because (1) Stephens didn’t indicate Godwin as the target (indeed, Stephens apparently picked Godwin at random), and (2) Stephens murdered Godwin “minutes” after making the post, and Facebook could not have intervened fast enough. Facebook could have notified the police in Stephens’ area (assuming Facebook knew where he was at that precise moment), but what could the police have done? In addition to the impossibly short turnaround time, they would not have had probable cause to arrest Stephens based on the Facebook post. At most, they could have surveilled Stephens, but that’s a likely misdirection of resources (diverting cops from dealing with actual crimes taking place) and there’s no guarantee that surveillance would have actually prevented the crime.

Another way of summarizing this is that Facebook almost certainly wasn’t a proximate cause of Godwin’s murder. I’m not sure why the court dismissed the case on duty rather than causation grounds, but the uncited Fields v. Twitter decision (and many similar ones) dismissed social media defendants from liability for actual terrorist attacks based on lack of proximate causation. Indeed, this case’s theory reminded me a little of the Seinfeld finale, which was premised on overexpansive legal theories of bystander liability.

I think the court used dry and technical tort analysis to sidestep these fact problems. In other words, even if the plaintiff got past the duty requirement, the case surely would fail on other grounds.

Analogy to Amazon Marketplace. The court doesn’t acknowledge the Stiner v. Amazon decision from the Ohio Supreme Court, which just came out a week beforehand. That ruling said Amazon wasn’t liable for the marketplace sale of pure caffeine, which a victim ingested and suffered a fatal overdose. Unlike Bolger v. Amazon from California, Amazon did not act as the fulfillment service. The Ohio Supreme Court held that Amazon wasn’t the “supplier” of the product, which negated strict liability and negligence. Like the Godwin opinion, the Stiner opinion is quite technical in nature and mostly ducked the policy issues; and like the Godwin opinion, there is a concurrence that blasts the tech industry. The Stiner concurrence writes: “The fact that the limited wording of the Act leaves a gap that allows e-commerce entities like Amazon to evade any obligation does not mean that there is a corresponding gap in the policy underlying the law; it means that the Act is failing to fully realize its purpose.” It would not surprise me if the Ohio legislature takes a greater interest in the concurrence.

[Note: the Stiner ruling could be quite significant nationally if it indicated that Amazon isn’t liable for marketplace sales. However, in light of the Bolger opinion, and Amazon’s willingness to accept statutory liability in California, I expect Amazon will choose to accept liability (so long as its competitors are equally affected) rather than double-down to preserve the reduced liability indicated by Stiner. Another possibility is that Amazon will try to make a bigger distinction between drop-shipped items, where Stiner reduces liability, and Fulfillment by Amazon, which the Bolger case indicated would create liability.]

So both the Stiner and Godwin opinions indicate that Ohio courts are imposing liability on tech companies narrowly, and only in accordance with the literal scope of legal doctrines. I’m a little surprised the Godwin opinion didn’t cite the Stiner case for that proposition, as it would have given greater cover to the general principles followed by the Godwin court.

The Concurrence’s Views Will Prevail. Despite its troubling nature, the concurrence previews our future. Regulators across the globe are seeking to impose broad-based duties of care on Internet services, such as the UK Online Harms proposal. Amidst the deluge of anti-Section 230 bills in Congress, several would add various duties of care.

One bill that stands out is the Manchin/Cornyn “See Something, Say Something Online Act of 2020,” which would require Internet services to notify the DOJ every time they know/should know about a “suspicious transmission” (defined as “any public or private post, message, comment, tag, transaction, or any other user-generated content or transmission that commits, facilitates, incites, promotes, or otherwise assists the commission of a major crime”). Imagine what this would look like: a flood of notices to the DOJ, many of which would be legally inconsequential, and all of which the DOJ would have to sort through. This mandatory reporting obligation creates obvious risks of weaponization and gaming, i.e., false accusations that could direct a life-changing FBI investigation towards innocent targets. The massive transfer of information to the DOJ also raises enormous privacy concerns. I, and I think many Americans, don’t want our country to constantly surveil our every move for potential illegality and unleash law enforcement based on merely suspicious behavior. That sounds less like America and more like Cuba’s “Committees for the Defense of the Revolution.”

Furthermore, the concurrence would allow plaintiffs to get discovery rather than grant the motion to dismiss. But with virtually every tort, victims could allege that the tortfeasor signaled his/her plans in social media posts and the social media service should have noticed and done something about those posts. I’m not sure the concurrence has really thought through the implications of this position, especially if the weak connections between Stephens’ post and the murder of Godwin are enough to survive a motion to dismiss. The flood of discovery would raise defense costs and create serious privacy issues. Unfortunately, many Section 230 reform bills would give plaintiffs more ways to get discovery, eviscerating Section 230’s procedural benefits.

In a couple of years, we’ll probably look back at this concurrence as prescient. Whether we’ll praise its prescience remains to be seen.

UPDATE: In February 2021, the Ohio Supreme Court denied a request to hear the case.

Case citation: Godwin v. Facebook, Inc., 2020 WL 5949862 (Ohio Ct. App. Oct. 8, 2020)