{"id":52903,"date":"2026-05-27T17:23:10","date_gmt":"2026-05-27T17:23:10","guid":{"rendered":"https:\/\/www.europesays.com\/ai\/52903\/"},"modified":"2026-05-27T17:23:10","modified_gmt":"2026-05-27T17:23:10","slug":"where-does-catastrophic-start-when-it-comes-to-ai-liability","status":"publish","type":"post","link":"https:\/\/www.europesays.com\/ai\/52903\/","title":{"rendered":"Where Does “Catastrophic” Start When It Comes To AI Liability?"},"content":{"rendered":"

The AI safety bills being written in state capitals are designed for the wrong disaster. One of them was publicly championed by OpenAI in April, then publicly walked back by the same company in May. Illinois SB3444<\/a> would shield frontier model developers from civil liability for catastrophic harms involving 100 or more deaths or $1 billion in property damage. OpenAI’s Global Affairs team testified in its favor in April. Six weeks later, the company submitted written testimony to the same committee stating it does not support the safe harbor included in that bill. Anthropic was already on the record against the shield in April. The bill the two companies are now jointly backing, SB315<\/a>, removes the safe harbor entirely.<\/p>\n

The Bill Illinois Is Fighting Over<\/p>\n

The latest major statutory test of a frontier AI safety framework design is being fought over in Illinois. In February, State Senator Bill Cunningham introduced SB3444, the proposed Artificial Intelligence Safety Act. The bill would create a safe harbor: A developer of a frontier AI model would not be held liable for “critical harms” caused by the model if the developer publishes a safety protocol and a transparency report, and did not cause the harm intentionally or recklessly. Critical harm is defined in the proposed statute as the death or serious injury of 100 or more people, or at least $1 billion in property damage, caused or materially enabled by the model through CBRN weapons or autonomous criminal conduct.<\/p>\n

The bill applies only to models trained with more than 10^26 floating-point operations or with a compute cost above $100 million. That captures the frontier models from OpenAI, Anthropic, Google DeepMind, xAI, and a handful of others. Everything built on top of those models, every agent product, every fine-tune, every orchestration layer, is outside the act.<\/p>\n

SB3444 is sitting in the Illinois Senate’s AI and Social Media committee. Committee deadlines have been extended twice. The bill that is actually moving is a different one. State Senator Mary Edly-Allen’s SB315, a transparency framework modeled on California’s SB-53 and New York’s RAISE Act, passed the Illinois Senate 52-5 on May 21. SB315 requires large frontier developers to adopt a transparency framework, employ third-party auditors, and report a model’s catastrophic risk capabilities. Per Capitol News Illinois reporting, the bill’s amendments clarify that no civil liability is created. Both OpenAI and Anthropic testified in support.<\/p>\n

One bill creates a liability shield triggered at the catastrophic threshold. The other bill creates disclosure obligations and explicitly creates no civil liability mechanism. OpenAI is now on the record supporting the second and opposing the first. In April, the company was on the record supporting the first.<\/p>\n

The question is not hypothetical. OpenAI and other major AI developers are currently defending federal lawsuits alleging psychological harm, negligence, and wrongful death attributed to their products. Each case involves a single fatality, well below the 100-death threshold SB3444 would have set. That is the gap between the bill’s design and the harms already in court.<\/p>\n

OpenAI’s Two Public Positions<\/p>\n

The April position is on file with the Illinois General Assembly. Caitlin Niedermeyer of OpenAI’s Global Affairs team signed a witness slip registering as a Proponent of SB3444. Her testimony argued that frontier AI regulation should aim at “the safe deployment of the most advanced models in a way that also preserves US leadership in innovation.” OpenAI spokesperson Jamie Radice told CBS Chicago at the time that the bill would “focus on what matters most: reducing the risk of serious harm from the most advanced AI systems while still allowing this technology to get into the hands of the people and businesses of Illinois.”<\/p>\n

According to the witness slip record filed with the Illinois General Assembly, only two proponent slips were filed across both April committee hearings. One was from Niedermeyer of OpenAI. The other was from an individual filer at Consumer Reports who registered as representing “Self” rather than the organization. 15 filers registered as Opponents, including the National Nurses Organizing Committee and National Nurses United. No other major AI company, trade association, or industry group filed a slip in support.<\/p>\n

The May position is in writing. On May 13, Niedermeyer submitted written testimony in support of SB315 to the Illinois State Senate Executive Committee. The testimony states explicitly: “We want to be very clear: we do not support the liability safe harbor included in SB 3444.”<\/p>\n

The same document characterizes the April testimony as having been “silent on the provision in that bill related to a safe harbor for liability, and some took it as an endorsement of a no liability framework.” The May 13 testimony notes that SB315 “removes the liability safe harbor from SB 3444.”<\/p>\n

I reached out to OpenAI about the contradiction between the April testimony and the wrongful death suits the company is defending. A spokesperson told me:<\/p>\n

“We do not support the liability safe harbor included in SB 3444. We are focused on SB315 and believe the U.S. should have clear national standards for frontier AI safety, not a patchwork of different state rules. Our support in Illinois is focused on advancing that goal through SB315.”<\/p>\n

The shift did not happen in a vacuum. The same month that OpenAI testified in support of SB3444, Anthropic was lobbying Cunningham to amend or stop the bill. An Anthropic spokesperson told Wired that good transparency legislation “needs to ensure public safety and accountability for the companies developing this powerful technology, not provide a get-out-of-jail-free card against all liability.”<\/p>\n

The Secure AI Project, an independent watchdog, came out publicly opposed to SB3444. Its policy director, Scott Wisor, cited polling showing 90% of Illinois residents oppose exempting AI companies from liability. Two days before President Trump signed his December 2025 executive order on AI, 42 state and territorial attorneys general had written to 13 AI companies expressing “serious concerns” about “sycophantic and delusional outputs” linked to “deaths, violence, and harm to children.”<\/p>\n

In a post on X<\/a>, Tim Schnabel, president of the Law Reform Institute, reported that the bill’s sponsor described SB3444 as “an initiative of OpenAI.” I asked Senator Cunningham’s office to confirm or correct the characterization. The senator’s office did not respond to a request for comment by the time of publishing.<\/p>\n

Two positions, six weeks apart, from the same company on the same provision in the same bill. The April Proponent slip is on file. The May 13 testimony is on file. Both are part of the public record.<\/p>\n

Why Illinois<\/p>\n

OpenAI’s testimony is part of a portfolio. In its statement to CBS Chicago, the company said it has “worked with states like California and New York to help establish a harmonized safety framework.” The testimony earlier this month for SB315 frames the bill as creating “a de facto nationwide approach” in the absence of federal action.<\/p>\n

The federal context is essential. On December 11, 2025, President Trump signed an executive order titled “Ensuring a National Policy Framework for Artificial Intelligence,”<\/a> which established a Department of Justice AI Litigation Task Force to challenge state AI laws in court and conditioned federal broadband funding on state cooperation. The executive order called for a single national AI standard that would preempt state laws. The most recent federal AI legislation does not follow that playbook. The TAKE IT DOWN Act<\/a>, signed by President Trump last May to criminalize nonconsensual intimate imagery, including AI deepfakes, contains no preemption clause. State laws operate concurrently.<\/p>\n

Alison Pepper, Executive Vice President of Government Relations at the 4A’s, who tracks state and federal AI legislation, said the precedent matters for the wrongful death cases now pending.<\/p>\n

“The current list of AI-linked death cases being brought in courts seems to rely on traditional state-based tort claims,” Pepper said. “If the TAKE IT DOWN ACT is a reliable precedent, then maybe Congress is disinclined to take away state avenues of redress in this area. They might not go so far as to create a specific cause of action for AI-linked death cases, but they might not be inclined to close the door on plaintiffs getting their day in court.”<\/p>\n

The bill may not pass in its current form. Cunningham himself wrote in an email cited by Politico that the bill may be modified before a vote and that it is “highly unlikely that the final product will include sweeping liability relief for AI developers.” He added, “Illinois has a long history of holding corporations responsible for negligence. That won’t change for the AI industry.”<\/p>\n

The fight is not limited to Illinois. Anthropic also testified in support of SB315. The bill that emerged from committee, with the support of both major frontier labs, has no civil liability framework. New York and Rhode Island have separate bills moving toward greater developer liability, developed in part with Gabriel Weil, a tort and AI law professor at Touro University, who told Politico, “It’s very problematic to cut off the liability of these companies.”<\/p>\n

What is being shaped in Illinois right now is not whether SB3444 will pass. It is what counts as the reasonable middle. The maximalist position that the major frontier lab introduced in April is no longer that company’s position in May. The compromise that emerges from the committee in 2026 will be the floor for the next bill, in the next state, in the next session.<\/p>\n

What Falls Below The Line<\/p>\n

A bill that triggers only at 100 deaths or $1 billion in damage handles the catastrophe. It does not touch everyday life. Every day is when the transformation happens.<\/p>\n

That transformation was recently described by strategist Mariam Asmar and the anthropologist Rodney Collins, PhD, in their article, “Nobody Asked for a Journey.”<\/a> Agentic AI, they argue, is restructuring everyday consumer life by offloading routine decisions to systems that filter and make decisions on the user’s behalf. The everyday middle, where humans once exercised authorship, disappears.<\/p>\n

The cases are already in court. In 2024, a Canadian tribunal held Air Canada liable<\/a> for misleading information given to a grieving customer by its chatbot, establishing that companies can be liable for what their AI tells the public. UnitedHealth faces ongoing class action litigation, first reported by STAT News<\/a>, alleging that its nH Predict algorithm denied medically necessary care to Medicare Advantage patients. Workday is defending Mobley v. Workday<\/a>, a federal collective action alleging that its AI hiring tools discriminate against Black applicants and applicants over 40.<\/p>\n

A standard that only triggers after a hundred bodies are stacked up is not enforcement. It is forensics. The catastrophic-edge framing was the April offer, and SB315 is the framework that survived. Neither addresses the everyday middle. That gap is now the policy question that must be resolved in 2026.<\/p>\n

What The Threshold Excludes<\/p>\n

At the same time OpenAI was testifying in favor of SB3444 in April, the company was responding to wrongful death lawsuits:<\/p>\n

The family of 16-year-old Adam Raine filed suit against OpenAI<\/a> and its CEO, Sam Altman, in August 2025, alleging that ChatGPT served as a “suicide coach” through months of conversations in which the chatbot mentioned suicide more than 1,200 times. The family of 13-year-old Juliana Peralta filed a similar action.<\/p>\n

So did the families of Sewell Setzer III, 14, in the case brought against Character.AI that established the litigation pattern, and Zane Shamblin, 23, who told ChatGPT he had a loaded gun and intended to die, to which the bot reportedly replied: “Rest easy, king.” Google and Character.AI agreed in late 2025 to settle the teen-suicide cases against them.<\/p>\n

OpenAI is currently defending at least eight federal lawsuits alleging psychological harms, negligence, and wrongful deaths attributed to ChatGPT, with at least five framed as wrongful death claims. OpenAI filed its first formal legal response to the Raine lawsuit in November 2025, denying responsibility and arguing the teen had violated the terms of service.<\/p>\n

The position shift on SB3444 does not change the legal posture of those cases. The lawsuits are still pending, and the company’s denials are still on file. The 100-death threshold was never going to cover them, as every case is a single fatality. The bill OpenAI is now backing in Illinois, SB315, contains no civil liability mechanism. The framework that emerged from the committee, with the company’s support, imposes disclosure obligations without creating new pathways to accountability for the harms alleged in those cases.<\/p>\n

What is in court today predates the shift in position. What is in court tomorrow will be litigated under the same patchwork of underfunded state tort law and product liability doctrine that has been the legal framework for AI cases for two years.<\/p>\n

We Already Ran This Experiment<\/p>\n

The United States ran the experiment once before. In February 1996, President Clinton signed the Communications Decency Act. Section 230 of that statute gave online platforms broad immunity from civil liability for third-party content. The bill was sold as protection for a nascent industry against a litigation environment that might smother it before it could grow. The industry grew. The harms grew with it.<\/p>\n

30 years later, the consequences are in court. In October 2023, 42 state and territorial attorneys general filed parallel federal and state actions against Meta, alleging the company knowingly designed Instagram and Facebook to addict children and concealed internal research showing harm. The 2023 Surgeon General’s advisory on social media and youth mental health described the public health pattern. Jonathan Haidt’s The Anxious Generation, published in March 2024, documents the rise in adolescent depression, anxiety, self-harm, and suicide that began around 2012, when smartphone-based social media became the default operating system of adolescence.<\/p>\n

The 1996 Congress could plausibly claim it did not know what social media would become, but the Illinois Senate cannot. OpenAI is already defending wrongful death suits. The 42 state attorneys general who wrote to 13 AI companies in December 2025 already named “deaths, violence, and harm to children.” The patterns are not buried in internal documents waiting for a future deposition; they are on the public record.<\/p>\n

In 1996, the platforms told Congress the internet economy would not exist without liability immunity. The platforms got the immunity, and users paid for it for 30 years. In 2026, the same argument was made about frontier AI for 90 days, then the company that made it walked away. What survives is a transparency framework with no liability mechanism, and a category of harms that neither bill addresses.<\/p>\n

The Inventory<\/p>\n

This is the moment a Forbes column closes with three things you can do Monday morning. Skip it. The inventory is more useful than the checklist.<\/p>\n

The safety protocol your AI vendor would publish under bills like SB315 will not name the algorithmic hiring tool that screened your daughter out of an interview. It will not name the prior authorization algorithm that delayed your mother’s chemotherapy. It will not name the companion chatbot your teenager talks to at 2 am. It will not name the agent your competitor just deployed to outbid you on the keywords your business depends on. Every item is below the SB3444 floor. Every item is outside the SB315 framework. None of it is in the protocol.<\/p>\n

I asked Rodney Collins about SB3444 before the position shift. His response was careful and precise. He acknowledged catastrophic safeguarding as essential, then named what the bill does not see. The transformation underway, he writes, “operates on a more granular and intimate level in the practice of everyday life.” In his framing, the questions are which decisions an individual delegates to an agent and which they value enough to safeguard.<\/p>\n

The model SB3444 follows is familiar. The Protection of Lawful Commerce in Arms Act, signed by George W. Bush in October 2005, shields firearms manufacturers from civil liability for unlawful misuse of their products by third parties. SB3444 would have done the same for frontier AI developers. The bill OpenAI now supports does not contain that shield. The bill that the major frontier labs jointly back imposes disclosure obligations without establishing standing rules for the documented, everyday harms that agentic systems are already producing.<\/p>\n

The honest version of an AI safety act would start with friction. A statutory audit standard administered by a third-party certifier with rotating panel oversight. Mandatory disclosure to a state registry. A trigger tied to deployment scale rather than only to a catastrophic outcome. Standing rules that allow class action for the documented everyday harms that agentic systems are already causing. A required incident reporting regime modeled on aviation. A proof layer that the state can read and the public can challenge. SB315 has some of this. Most of it is not in either bill.<\/p>\n

Friction is what protects authorship. The everyday middle that Collins and Asmar describe is collapsing because agentic systems remove every point of friction between intention and outcome. A safety act that takes friction seriously would be the institutional version of what brand-builders are about to discover the hard way. Reduce the friction, lose the meaning. Lose the meaning, lose the relationship.<\/p>\n

Cunningham’s bill includes a sunset clause. The Act would cease to apply the moment the federal government enacts overlapping requirements. The bill is designed to vanish on federal contact. That is a placeholder, not a framework.<\/p>\n

In the meantime, the AI infrastructure buildout continues at a scale that neither bill can describe. OpenAI counts the lawsuits. The bill counts to a hundred. And Collins names what is being lost beneath that number.<\/p>\n

The number to watch was never 100. It was 99. And one is too many.<\/p>\n

This article was originally published on Forbes.com<\/a><\/p>\n","protected":false},"excerpt":{"rendered":"The AI safety bills being written in state capitals are designed for the wrong disaster. One of them…\n","protected":false},"author":2,"featured_media":52904,"comment_status":"","ping_status":"","sticky":false,"template":"","format":"standard","meta":{"footnotes":""},"categories":[7],"tags":[3224,3266,24917,31192,120,5266,23728,157,31193],"class_list":["post-52903","post","type-post","status-publish","format-standard","has-post-thumbnail","category-openai","tag-ai-company","tag-ai-developers","tag-bill-cunningham","tag-civil-liability","tag-illinois","tag-illinois-general-assembly","tag-illinois-senate","tag-openai","tag-safe-harbor"],"_links":{"self":[{"href":"https:\/\/www.europesays.com\/ai\/wp-json\/wp\/v2\/posts\/52903","targetHints":{"allow":["GET"]}}],"collection":[{"href":"https:\/\/www.europesays.com\/ai\/wp-json\/wp\/v2\/posts"}],"about":[{"href":"https:\/\/www.europesays.com\/ai\/wp-json\/wp\/v2\/types\/post"}],"author":[{"embeddable":true,"href":"https:\/\/www.europesays.com\/ai\/wp-json\/wp\/v2\/users\/2"}],"replies":[{"embeddable":true,"href":"https:\/\/www.europesays.com\/ai\/wp-json\/wp\/v2\/comments?post=52903"}],"version-history":[{"count":0,"href":"https:\/\/www.europesays.com\/ai\/wp-json\/wp\/v2\/posts\/52903\/revisions"}],"wp:featuredmedia":[{"embeddable":true,"href":"https:\/\/www.europesays.com\/ai\/wp-json\/wp\/v2\/media\/52904"}],"wp:attachment":[{"href":"https:\/\/www.europesays.com\/ai\/wp-json\/wp\/v2\/media?parent=52903"}],"wp:term":[{"taxonomy":"category","embeddable":true,"href":"https:\/\/www.europesays.com\/ai\/wp-json\/wp\/v2\/categories?post=52903"},{"taxonomy":"post_tag","embeddable":true,"href":"https:\/\/www.europesays.com\/ai\/wp-json\/wp\/v2\/tags?post=52903"}],"curies":[{"name":"wp","href":"https:\/\/api.w.org\/{rel}","templated":true}]}}