Cycle on May 24, 2026

I will begin by listing the open threads of interest for this topic to understand what specific areas or questions need to be pursued in this cycle.

I will search the web for information regarding the federal stay in `xAI LLC v. Philip J. Weiser` and Colorado's AI Act rulemaking or legislative updates (such as SB 24-205 or SB 26-189).

Now I will fetch some of these highly relevant URLs to understand the details of `xAI LLC v. Philip J. Weiser` and Colorado's AI Act litigation. Let's fetch the Baker Botts, Norton Rose Fulbright, and Statescoop articles, and if possible, we should also look at the Troutman Privacy PDF minute order. Let's start with the Norton Rose Fulbright article because it seems to summarize the lawsuit, the DOJ intervention, and the suspension of enforcement. Let's also fetch the Statescoop article. Let's fetch the Baker Botts article too. Let's start with Norton Rose Fulbright.

This is incredibly interesting! Let's examine the details of the lawsuit: 1. `xAI LLC v. Philip J. Weiser` (Case No. 1:26-cv-01515-DDD-CYC in the District of Colorado). 2. The lawsuit challenges SB 24-205 (the Colorado AI Act, which was signed into law on May 17, 2024, and was scheduled to go into effect on June 30, 2026). Note: The thread description mentions "SB 26-189" and "rulemaking process for SB 26-189 (which must be completed on or before January 1, 2027)". Let's investigate whether SB 26-189 is a newly proposed or enacted bill that amended or replaced SB 24-205, or if there is a different bill. 3. The US DOJ intervened on April 24, 2026, to oppose the law. Wait! The DOJ intervened in the case. Let's find out why the US Department of Justice intervened to *oppose* the law, or if it intervened as a plaintiff-intervenor or defendant-intervenor, or what exactly its position is. The Norton Rose Fulbright article says: "On April 24, 2026, the US Department of Justice moved to intervene in the case to also oppose the law." Let's check the complaint in intervention or other sources to confirm. 4. On April 27, 2026, the court granted a joint motion by X.AI and the Colorado AG to suspend enforcement of the AI Act until: - the 2026 Regular Session of the Colorado General Assembly and any subsequent rulemaking related to the AI Act concludes, and - X.AI's forthcoming motion for preliminary injunction is decided. Let's search for "SB 26-189" or "SB26-189" to see what that bill is. Is it a bill introduced in the 2026 Colorado legislative session? Let's do a web search.

Wow! This is a massive development. Let's fetch several of these resources. 1. `https://www.reedsmith.com/our-insights/blogs/technology-law-dispatch/102mrzf/sb-26-189-colorado-legislature-kicks-off-caia-rewrite-race/` (Reed Smith - SB 26-189: Colorado legislature kicks off CAIA rewrite race) 2. `https://www.consumerfinancemonitor.com/2026/05/12/colorado-rewrites-its-landmark-ai-law-unpacking-sb-26-189-and-what-it-means-for-businesses/` (Consumer Finance Monitor - Colorado rewrites its landmark AI law: Unpacking SB 26-189 and what it means for businesses) 3. `https://www.buchalter.com/insights/colorado-rewrites-its-ai-law-what-employers-must-know-about-sb-26-189/` (Buchalter - Colorado Rewrites Its AI Law: What Employers Must Know About SB 26-189) 4. `https://www.clarkhill.com/news-events/news/colorado-ai-act-update-sb26-189-key-changes-2027/` (Clark Hill - Colorado AI Act Update: Key Changes in SB26-189, New in 2027) Let's fetch the Consumer Finance Monitor article first. Then we'll fetch the Reed Smith and Buchalter ones. This is extremely fresh and directly answers the thread about Colorado's AI Act rewrite and SB 26-189! Let's start with Consumer Finance Monitor.

This is a massive development. Let's look at the other articles to gather more specific details, quotes, and context. Let's fetch the Buchalter article: `https://www.buchalter.com/insights/colorado-rewrites-its-ai-law-what-employers-must-know-about-sb-26-189/` And the Clark Hill article: `https://www.clarkhill.com/news-events/news/colorado-ai-act-update-sb26-189-key-changes-2027/` And the Reed Smith article: `https://www.reedsmith.com/our-insights/blogs/technology-law-dispatch/102mrzf/sb-26-189-colorado-legislature-kicks-off-caia-rewrite-race/` Let's fetch Buchalter first.

This is absolutely brilliant. It provides a complete, clear, and extremely detailed update on the Colorado AI Act rewrite, SB 26-189, the lawsuit `xAI LLC v. Philip J. Weiser`, and how the DOJ's intervention and the court-ordered stay fit together. Let's fetch the Clark Hill article as well to see if there are additional details: `https://www.clarkhill.com/news-events/news/colorado-ai-act-update-sb26-189-key-changes-2027/` And let's fetch the Reed Smith article too: `https://www.reedsmith.com/our-insights/blogs/technology-law-dispatch/102mrzf/sb-26-189-colorado-legislature-kicks-off-caia-rewrite-race/` Let's fetch Clark Hill first.

Let's fetch the Reed Smith article to get even more legal perspective on the "CAIA rewrite race": `https://www.reedsmith.com/our-insights/blogs/technology-law-dispatch/102mrzf/sb-26-189-colorado-legislature-kicks-off-caia-rewrite-race/`

This is fantastic. We now have a highly detailed understanding of Colorado SB 26-189 (the "new" Colorado AI Act, which repeals and replaces the original SB 24-205). Let's double-check the timelines: - SB 24-205 was originally set to take effect on June 30, 2026 (after being delayed from Feb 1, 2026). - In early April 2026, xAI LLC (developer of Grok) filed a lawsuit against Colorado AG Philip J. Weiser in the federal District of Colorado (`xAI LLC v. Philip J. Weiser`, Case No. 1:26-cv-01515-DDD-CYC). - On April 24, 2026, the US DOJ moved to intervene in the case to oppose the law, acting under the executive branch's recent directive (under Executive Order 14365) to challenge state AI regulations. - On April 24, 2026, xAI and the Colorado AG filed a joint motion to stay/suspend enforcement of the AI Act until the legislative session and subsequent rulemaking concluded, and xAI's motion for preliminary injunction was decided. - On April 27, 2026, the court granted this joint motion, temporarily staying enforcement of SB 24-205. - On May 14, 2026, Colorado Governor Jared Polis signed SB 26-189 into law. - SB 26-189 completely repeals and replaces SB 24-205. It replaces the broad "high-risk AI" and "algorithmic discrimination" framework with a narrower framework focused on "automated decision-making technology" (ADMT) that processes personal data to "materially influence" a "consequential decision." - The new law takes effect on January 1, 2027. - However, the litigation stay remains in effect, and the Colorado AG has stated he does not intend to enforce SB 24-205 or any replacement legislation (including SB 26-189) until after the mandatory AG rulemaking process has concluded. - The AG's rulemaking must be completed on or before January 1, 2027. - SB 26-189 eliminates several of the most burdensome requirements of the original law (such as mandatory NIST AI RMF/ISO 42001 risk management programs, annual impact assessments, self-reporting of algorithmic discrimination, and a freestanding duty of reasonable care), but it retains and tightens specific consumer-facing obligations like point-of-interaction notices, post-adverse-outcome disclosures within 30 days, 3-year recordkeeping, and a demanding right to "meaningful human review" and reconsideration. - It also introduces a comparative fault framework between developers and deployers, explicitly voiding contractual provisions that attempt to shield a party from its own discriminatory acts, and directly addressing liability questions raised in cases like `Mobley v. Workday`. - In addition, SB 26-189 eliminated some conditional federal exemptions for certain federally regulated entities that were in the original SB 24-205, meaning more entities are now in scope. Let's check the status of the other thread: "German Federal Court (BGH) Appeal on Chatbot Hallucination Liability" (OLG Hamm decision, case No. 4 UKl 3/25). Let's search the web for "OLG Hamm" "4 UKl 3/25" OR "BGH" "chatbot" "hallucination" OR "unfair competition" "hallucination" "BGH". Let's run a search.

This is incredible! The OLG Hamm decision was issued on **May 12, 2026** (case No. I-4 UKl 3/25 or 4 UKl 3/25). Let's fetch the SKW Schwarz article, the Beckmann und Norda article, and the Heise Online article to get all the details of this landmark decision. Let's do a search or fetch these directly. Let's fetch: 1. `https://www.skwschwarz.de/news/haftung-fur-irrefuhrende-aussagen-eines-ki-chatbots` (SKW Schwarz - Haftung für irreführende Aussagen eines KI-Chatbots) 2. `https://www.heise.de/news/Dr-Rick-Dr-Nick-Aerzte-verlieren-wegen-KI-Halluzinationen-vor-Gericht-11293866.html` (Heise Online - Firma haftet für ihren Chatbot) 3. `https://moselwal.com/blog/olg-hamm-ai-liability-architecture-dialog-systems` (Moselwal - OLG Hamm 4 UKl 3/25 — AI chatbot liability & architecture) Let's fetch SKW Schwarz first.

This is an exceptional find! Let's fetch the Heise Online article as well to confirm other details about the case: `https://www.heise.de/news/Dr-Rick-Dr-Nick-Aerzte-verlieren-wegen-KI-Halluzinationen-vor-Gericht-11293866.html`

This is fantastic! Now let's do a search for other major global AI liability developments. Let's see: are there other court decisions, enforcement actions, legislative proposals, or regulatory updates that occurred in May 2026? Let's search for "AI liability" OR "AI regulation" OR "AI lawsuit" with `recent=true` to see what has happened in the last week. Or search for general May 2026 developments. Let's do a search.

Let's do a search on recent cases or developments in the US or EU. In the previous cycle summaries: - The EU AI Liability Directive was withdrawn in October 2025, and the Revised Product Liability Directive (PLD) was adopted. - Case "Mobley v. Workday" was mentioned, which is a major case in employment AI liability. Let's see if there are any recent updates on `Mobley v. Workday` in May 2026. Let's search.

This is phenomenal! The `Mobley v. Workday` case has had major recent developments in late April and May 2026. Let's fetch some of these articles to understand what happened. Let's fetch: 1. `https://www.akingump.com/en/insights/ai-law-and-regulation-tracker/court-allows-discrimination-claims-against-ai-hiring-tool-to-proceed-or-mobley-v-workday-inc` (Akin Gump - Court Allows Discrimination Claims Against AI Hiring Tool to Proceed | Mobley v. Workday, Inc.) 2. `https://www.maynardnexsen.com/publication-emerging-liability-for-ai-driven-hiring-tools-key-developments-in-mobley-v-workday-inc` (Maynard Nexsen - Key Developments in Mobley v. Workday, Inc.) 3. `https://www.hrexaminer.com/p/what-the-court-actually-did-in-mobley` (HRExaminer - What the Court Actually did in Mobley) Let's fetch Akin Gump first.

The Akin Gump summary is a bit short. Let's fetch the Maynard Nexsen article to see more details and dates: `https://www.maynardnexsen.com/publication-emerging-liability-for-ai-driven-hiring-tools-key-developments-in-mobley-v-workday-inc`

This is extremely helpful. Let's look at the dates and details: - The case is `Derek L. Mobley v. Workday, Inc.`, Case No. 3:23-cv-00770-RFL (N.D. Cal.). - In early 2026 (or late 2025/early 2026), there have been some key developments, including conditional certification of an age-based collective, and the court allowing claims to proceed on the basis that Workday is an "agent" of its client-employers and therefore falls within the statutory definition of "employer" under Title VII, the ADEA, and the ADA. - Wait, the Maynard Nexsen article mentions: "The Executive Branch’s stance on AI and disparate impact litigation changed between the Biden and Trump administrations... on April 23, 2025, Executive Order 14281 was issued, which instructs federal agencies to eliminate or reduce the pursuit of disparate impact theories of liability. The EEOC has since deprioritized disparate impact cases, such as the theories presented in Mobley. However, individual litigants are not restricted from pursuing such claims." - This is an extremely valuable piece of context! It shows the regulatory and administrative landscape shifting (deprioritization of disparate impact cases by the EEOC under EO 14281, while individual litigants continue to pursue these claims in court). - Wait, what about the recent order in `Mobley v. Workday` in early 2026? Let's check the HR Dive article: "Workday plaintiffs submit amended complaint reupping..." and other articles to see what happened in March or April 2026. Let's do a search.

Excellent! Let's fetch the HRExaminer article and the HR Dive article to see the exact details of the March 6, 2026, order in `Mobley v. Workday` and the subsequent developments. Let's fetch HRExaminer first, then HR Dive. This is very fresh and directly relevant to the developer-deployer liability split and how courts handle AI hiring screening tools. Let's fetch HRExaminer.

Let's fetch the HR Dive article too to see what else they say about the March 6, 2026, order or other recent actions: `https://www.hrdive.com/news/workday-partial-loss-judge-refuses-claims-dismissal/814227/`

This is extremely rich! Let's examine the details: - On **March 6, 2026**, Judge Rita F. Lin of the Northern District of California issued a major order in `Derek L. Mobley v. Workday, Inc.`, 3:23-cv-00770-RFL. - The order was a split ruling on Workday's motion to dismiss. - **The Wins for Plaintiffs:** - The court rejected Workday's argument that the ADEA (Age Discrimination in Employment Act) does not permit disparate impact claims by job applicants. Workday had argued that the Supreme Court's *Loper Bright* ruling (which ended *Chevron* deference) should invalidate prior district court precedent on this point. - The court held that *Loper Bright* did not apply because the prior precedent did not rely on *Chevron* deference, and that the EEOC's longstanding interpretation (that the ADEA applies to applicants) is still persuasive under *Skidmore* deference. - **The Wins for Workday:** - The court dismissed the California state law (FEHA) claims of non-California plaintiffs because the complaint lacked factual allegations tying them to California. - The court dismissed the ADA (disability) claim of one plaintiff (a cancer survivor with asthma) because the complaint failed to allege facts showing how Workday's automated systems could detect or discriminate against physical disabilities like cancer or asthma. - **The Next Steps / Amending:** - The court granted plaintiffs leave to amend by March 27, 2026. - The plaintiffs filed their amended complaint on or around March 27, 2026, reupping the physical disability and state-law claims with additional factual allegations. - **The Broader Legal Framework:** - This case is a "defining moment" for AI hiring tools, establishing that software vendors like Workday can be held liable as "agents" of employers under federal anti-discrimination laws. - It also mentions a second prominent California lawsuit against **Eightfold AI Inc.**, which allegedly used AI to build reports on job applicants without their knowledge or consent, ranking candidates based on their "likelihood of success" using third-party sources. Let's research this case! Let's do a search for "Eightfold AI" lawsuit 2026 to see what this case is about and if there are any recent updates. This will be an excellent additional finding.

This is an absolutely fascinating development! The lawsuit against Eightfold AI was filed on **January 20, 2026** in California. Let's fetch: 1. `https://www.workforcebulletin.com/ai-hiring-tools-and-consumer-reports-understanding-the-eightfold-litigation` (Workforce Bulletin - AI Hiring Tools and Consumer Reports: Understanding the Eightfold Litigation) 2. `https://www.insidetechlaw.com/blog/2026/03/class-action-questions-whether-using-ai-to-score-job-applicants-violates-the-fcra` (Inside Tech Law - Class action questions whether using AI to “score” job applicants violates the FCRA) 3. `https://www.cdflaborlaw.com/blog/ai-lawsuit-pushes-the-boundaries-of-ai-litigationand-may-signal-a-new-wave` (CDF Labor Law - AI Lawsuit Pushes the Boundaries of AI Litigation—and May Signal a New Wave) Let's fetch Workforce Bulletin first.

This is incredibly useful! Let's fetch the CDF Labor Law article to see if it adds more details: `https://www.cdflaborlaw.com/blog/ai-lawsuit-pushes-the-boundaries-of-ai-litigationand-may-signal-a-new-wave`

Ah, the CDF Labor Law article didn't have much content in its fetched markdown. Let's fetch the Inside Tech Law article: `https://www.insidetechlaw.com/blog/2026/03/class-action-questions-whether-using-ai-to-score-job-applicants-violates-the-fcra`

This is an incredibly rich, unique, and fresh finding. Let's look at the details: - **Case:** `Erin Kistler and Sruti Bhaumik v. Eightfold AI Inc.`, filed on **January 20, 2026**, in the Superior Court of California, County of Contra Costa (Case No. C26-00214). - **The Core Claim:** Unlike other AI hiring lawsuits that focus on discrimination or bias (like `Mobley v. Workday`), this class action seeks to establish that using AI-powered candidate ranking and scoring tools constitutes a violation of the **Federal Fair Credit Reporting Act (FCRA)** and **California's Investigative Consumer Reporting Agencies Act (ICRAA)**. - **Why this is a game-changer:** If the court rules that AI-generated candidate evaluations and "likelihood of success" scores (on a 0-5 scale) constitute "consumer reports" (or "credit reports") under the FCRA/ICRAA, then: 1. AI vendors like Eightfold AI will be legally classified as **Consumer Reporting Agencies (CRAs)**. 2. Vendors would be subject to strict statutory requirements: verifying employer certifications, ensuring maximum possible accuracy, providing report copies to consumers upon request, and offering dispute/correction procedures. 3. Employers using these tools would have to provide standalone disclosures, obtain prior written authorization from applicants, and provide pre-adverse action notices with copies of the AI scores/reports *before* rejecting candidates. - **The Allegations:** The plaintiffs claim that Eightfold's "Match Score" algorithm uses a proprietary LLM trained on 1.5 billion global data points (scraping career trajectories, skills, professional history from blogs, publications, etc.) to generate automated evaluations. Lower-ranked candidates are allegedly filtered out automatically before any human review, and candidates are never notified, given a chance to consent, or allowed to dispute inaccurate data. Let's check the status of the "German Federal Court (BGH) Appeal on Chatbot Hallucination Liability" (OLG Hamm decision, case No. 4 UKl 3/25, May 12, 2026). Wait! The OLG Hamm decision was issued on May 12, 2026. The court ruled that companies are fully liable under § 5 of the German Unfair Competition Act (UWG) for misleading or false statements made by their AI chatbots (hallucinations), even if the chatbot was trained on correct data and the false statement was purely an AI hallucination. This is because: 1. The chatbot is not a "third party" but an integrated part of the company's business operations. 2. The company is directly responsible (unmittelbar zuzurechnen) for the chatbot's public communications. 3. The court allowed an appeal (Revision) to the German Federal Court of Justice (BGH) due to the fundamental legal significance of the attribution of AI-generated statements. This is a perfect resolution to the German Federal Court (BGH) Appeal thread! Let's write down these findings. Wait, are there other major global AI liability developments we can find? Let's see if there are any other recent court decisions, enforcement actions, or regulatory updates. Let's search for "AI liability" EU 2026 or "Revised Product Liability Directive" 2026. Let's do a search.

This is very interesting too! Let's fetch the Freshfields article: `https://www.freshfields.com/en/our-thinking/blogs/risk-and-compliance/product-risks-today-how-the-new-product-liability-directive-turns-ai-act-complia-102mpu2` (Freshfields - Product Risks Today: How the new Product Liability Directive turns AI Act compliance into a question of liability, April 21, 2026) And let's fetch the Gibson Dunn article: `https://www.gibsondunn.com/eu-product-liability-directive-responding-to-software-ai-and-complex-supply-chains/` (Gibson Dunn - EU Product Liability Directive: Responding to Software, AI and Complex Supply Chains, March 23, 2026) Let's fetch Freshfields first. This will help us understand the EU side of AI liability and how the new Product Liability Directive (PLD) intersects with the EU AI Act as the December 9, 2026 transposition deadline approaches.

This is an absolute masterpiece of an article! It explains exactly how the new EU Product Liability Directive (PLD) creates a direct link between compliance with the EU AI Act and strict civil liability. Let's synthesize the key points from this: 1. **Transposition Deadline:** EU Member States have until **December 9, 2026** to transpose the revised Product Liability Directive (PLD) into national law. 2. **AI explicitly classified as a "product":** Under Article 4(1) PLD, software, including AI systems (whether cloud-based, on-device, or Software-as-a-Service), explicitly qualifies as a "product" subject to strict liability. 3. **The "Gateway" to Strict Liability (Article 7(1)):** A product is deemed "defective" under the PLD if it does not provide the safety that a person is entitled to expect or that is required under Union or national law. This directly links civil liability to regulatory compliance. Non-compliance with the EU AI Act (e.g., accuracy, robustness, cybersecurity, data governance, and transparency rules) will be treated by courts as a de facto benchmark for determining "defectiveness" under product liability law. 4. **Abandonment of the "Factory Gate Principle" (Article 7(2)(e)):** Traditionally, defectiveness was assessed at the moment a product was placed on the market. Under the new PLD, because AI systems continuously learn, receive updates, and are subject to patches, manufacturers are liable for defects arising *after* market placement if they retain control (e.g., through over-the-air updates or SaaS delivery). This creates a dynamic, ongoing obligation to maintain compliance with safety standards. 5. **Rebuttable Presumptions of Defectiveness (Article 10):** - If a claimant demonstrates non-compliance with a mandatory product safety requirement (such as the AI Act's risk management or cybersecurity standards), the product's defectiveness is legally presumed. - If a defendant fails to comply with a court order to disclose "relevant evidence" (e.g., design docs, training data, source code), defectiveness is presumed. - For cases of "scientific and technical complexity" (which AI systems represent), courts may presume both defect and causation if the claimant shows they are likely. This effectively shifts the burden of proof to the AI manufacturer. Let's review the threads. 1. Thread 1: "Colorado AG Rulemaking and Federal Stay Status for SB 26-189" - Status: We have fully investigated this. - Findings: - SB 24-205 was slated to take effect June 30, 2026. - In early April 2026, xAI LLC sued Colorado AG Philip J. Weiser in federal court (`xAI LLC v. Philip J. Weiser`, Case No. 1:26-cv-01515-DDD-CYC). - On April 24, 2026, the US DOJ intervened in support of xAI (marking a major first for the DOJ challenging a state AI law, stemming from Executive Order 14365). - On April 27, 2026, the court granted a joint motion to temporarily suspend enforcement of the AI Act. - On May 14, 2026, Colorado Governor Jared Polis signed SB 26-189, which completely repeals and replaces SB 24-205. - SB 26-189 shifts the focus from broad "high-risk AI" to "covered ADMT" (automated decision-making technology) materially influencing "consequential decisions." - SB 26-189 eliminates several burdensome compliance requirements (like mandatory NIST AI RMF programs, annual impact assessments, and a freestanding duty of reasonable care), but retains and tightens consumer-facing rights (notices, 30-day post-adverse-outcome disclosures, 3-year recordkeeping, and a strict "meaningful human review" standard). - It also eliminates some conditional federal exemptions, bringing more federally regulated entities into scope. - The law takes effect on January 1, 2027. - Rulemaking must be completed on or before January 1, 2027. - The AG has stated he does not intend to enforce the law until after rulemaking concludes, meaning enforcement of the new SB 26-189 is also effectively on hold during this rulemaking sprint. 2. Thread 2: "German Federal Court (BGH) Appeal on Chatbot Hallucination Liability" - Status: We have fully investigated this. - Findings: - On May 12, 2026, the Higher Regional Court of Hamm (Oberlandesgericht Hamm - OLG Hamm) issued its landmark ruling (Case No. I-4 UKl 3/25). - The court ruled that companies are strictly liable under § 5 of the German Unfair Competition Act (UWG) for false or misleading statements made by their AI chatbots (hallucinations), even if the company did not provide any incorrect data to the chatbot and the hallucination was entirely a system error. - The court established that a chatbot is not a "third party" but an integrated part of the company's business operations, making its communications directly attributable (unmittelbar zuzurechnen) to the company. - The case involved "Dr. Rick & Dr. Nick" (Aesthetify GmbH), whose chatbot falsely claimed they held non-existent specialized medical titles ("Fachärzte für ästhetische Medizin"). - The OLG Hamm admitted an appeal (Revision) to the German Federal Court of Justice (BGH) due to the fundamental significance of the attribution of AI-generated statements, making this the most highly anticipated AI liability case in German history. Let's write these up as comprehensive findings using the `write_note` tool. Wait, let's verify if there are any other findings from our search. We also found a third incredible development: 3. **The Eightfold AI Class Action (FCRA and ICRAA Liability):** - On January 20, 2026, two job applicants filed a class action lawsuit against Eightfold AI Inc. (`Erin Kistler and Sruti Bhaumik v. Eightfold AI Inc.`, Case No. C26-00214) in Contra Costa County, California. - The lawsuit argues that using AI-powered candidate evaluation and "Match Score" systems to automatically screen and rank applicants constitutes generating undisclosed, illegal "consumer reports" (or "credit reports") under the Federal Fair Credit Reporting Act (FCRA) and California's Investigative Consumer Reporting Agencies Act (ICRAA). - This is a major shift in AI litigation: rather than focusing on discrimination, it targets the lack of compliance with consumer credit reporting rules (verification of certifications, disclosures, prior written consent, and dispute/correction rights). - If successful, it would classify AI hiring vendors as Consumer Reporting Agencies (CRAs), imposing massive administrative and compliance burdens on both vendors and employers. We also have a fourth major development: 4. **The `Mobley v. Workday` March 6, 2026 Order and Subsequent Developments:** - In `Derek L. Mobley v. Workday, Inc.`, Case No. 3:23-cv-00770-RFL (N.D. Cal.), Judge Rita F. Lin issued a split ruling on March 6, 2026. - The court ruled that job applicants can bring disparate impact age discrimination claims under the ADEA, rejecting Workday's argument that the end of *Chevron* deference (under *Loper Bright*) invalidated prior precedent. The court found that *Skidmore* deference still made the EEOC's interpretation persuasive. - The court dismissed some state-law and physical disability claims but allowed plaintiffs to amend. On March 27, 2026, plaintiffs submitted their amended complaint. - This case is a "defining moment" establishing that AI vendors can be held liable as "agents" of employers under federal anti-discrimination laws. We also have a fifth major development: 5. **The EU Product Liability Directive (PLD) and AI Act Strict Liability Link:** - As the December 9, 2026 transposition deadline approaches, the revised EU Product Liability Directive (PLD) explicitly classifies software and AI systems as "products" subject to strict liability. - It links civil liability directly to regulatory compliance: non-compliance with the EU AI Act (cybersecurity, accuracy, data governance, transparency) will be treated by courts as a de facto benchmark for determining "defectiveness" under product liability law. - It abandons the "factory gate principle," making manufacturers liable for defects arising after market placement due to continuous learning and over-the-air updates. - It introduces powerful rebuttable presumptions of defectiveness (for non-compliance with safety rules, non-compliance with court-ordered disclosures, or due to "scientific and technical complexity" of AI), effectively shifting the burden of proof to the AI manufacturer. Let's write these findings as topic notes using `write_note`. Wait! We should check if we can reuse any existing slugs. The auto-injected findings include: - `global-ai-enforcement-q1-2026-trends` (Global AI Enforcement Landscape Q1 2026: Data Privacy, AI-Washing, and Operational Risk) - `cycle-summary-2026-05-20` (May 20, 2026 Cycle Summary: Global AI Liability Developments) - `cycle-summary-2026-05-23` (May 23, 2026 Cycle Summary: Global AI Liability, Contractual Gaps, and Regulatory Resets) - `italy-court-of-pistoia-ai-unfair-competition-liability-2026` (Italy's Court of Pistoia: AI-Generated Content Does Not Exempt Deployers From Unfair Competition Liability (March 2026)) - `uk-eu-us-ai-governance-convergence-may-2026` (UK, EU, and US Regulatory Convergence: AI Governance as Compliance Emergency (May 2026)) Wait! Since these are fresh findings, let's write them using fresh, descriptive slugs. Let's write four major substantive findings and one cycle summary finding: 1. Finding 1: Colorado's AI Act Reset: SB 26-189 Repeals SB 24-205 Amidst Federal Stay in `xAI v. Weiser` - Slug: `colorado-ai-act-sb26-189-repeals-sb24-205-federal-stay-2026` 2. Finding 2: German Higher Regional Court (OLG Hamm) Rules Chatbot Hallucinations Trigger Strict Unfair Competition Liability - Slug: `germany-olg-hamm-chatbot-hallucination-strict-liability-2026` 3. Finding 3: AI Hiring Under Fire: Eightfold AI Class Action Seeks to Classify Candidate Ranking as FCRA Consumer Report - Slug: `eightfold-ai-class-action-fcra-consumer-reporting-agency-2026` 4. Finding 4: Mobley v. Workday: March 2026 Order Reaffirms ADEA Applicant Coverage and Vendor Agency Status Post-Chevron - Slug: `mobley-v-workday-march-2026-order-adea-vendor-agency-liability` 5. Finding 5: EU Product Liability Directive: Transposition Countdown Links AI Act Non-Compliance to Strict Liability - Slug: `eu-product-liability-directive-pld-ai-act-strict-liability-2026` Wait, let's also write a cycle-summary note for today, May 24, 2026. - Slug: `cycle-summary-2026-05-24` Let's write each note with high precision, naming and linking the actual things, and carrying 1-2 representative verbatim quotes with their links, as required by the developer guidelines. Let's draft Note 1: Colorado's AI Act Reset. Wait, let's check the exact details of the quotes and links we will use. From `https://www.buchalter.com/insights/colorado-rewrites-its-ai-law-what-employers-must-know-about-sb-26-189/`: "On May 14, 2026, Governor Polis signed Senate Bill 26-189 into law, repealing and replacing Colorado’s 2024 landmark AI law... Senate Bill 26-189 — A Bill Concerning the Use of Automated Decision-Making Technology in Consequential Decisions (“ADMT”) — supersedes SB 24-205 and reshapes the compliance landscape for employers using AI in hiring, compensation, and workforce management." "Critically, the Attorney General has stated he does not intend to enforce SB 24-205 or any legislation replacing or amending it — including SB 26-189 — until after the rulemaking process has concluded." From `https://www.nortonrosefulbright.com/en-us/knowledge/publications/de3ad9de/xai-sues-doj-intervenes-enforcement-of-colorado-ai-act-suspended`: "On April 24, 2026, the US Department of Justice moved to intervene in the case to also oppose the law. Notably, this is the first time that the DOJ has sought to intervene in a lawsuit challenging a state AI law..." "On April 27, 2026, the Court granted the Parties’ joint motion, temporarily suspending enforcement of the AI Act." Let's call `write_note` for Note 1.

I will write the topic note regarding Colorado's AI Act Reset (SB 189 repealing SB 205, the federal stay, and DOJ's intervention in `xAI LLC v. Philip J. Weiser`).

I will write the second topic note detailing the landmark German Higher Regional Court (OLG Hamm) decision on strict unfair competition liability for chatbot hallucinations.

I will write the third topic note regarding the newly filed class action against Eightfold AI, which seeks to classify AI-driven candidate scoring and ranking as undisclosed "consumer reports" under the federal Fair Credit Reporting Act (FCRA).

I will write the fourth topic note regarding the March 6, 2026, order in `Mobley v. Workday, Inc.`, which permitted disparate impact age discrimination claims to proceed and reaffirmed the vendor's potential "agent" liability under traditional federal civil rights laws.

I will write the fifth topic note regarding the EU Product Liability Directive (PLD) and how its upcoming December 9, 2026 transposition deadline effectively transforms EU AI Act compliance into a question of civil strict liability.

I will write the final cycle summary note for May 24, 2026, which synthesizes these five major developments in global AI liability.

I have successfully written all five substantive findings and the cycle summary. I will now resolve the two open threads that have been thoroughly addressed in this cycle. 1. Thread `019e53e0-8a5e-78ce-a461-fe23ab218737` (Colorado AG Rulemaking and Federal Stay Status for SB 26-189) 2. Thread `019e53e0-7e06-7997-a980-b79510df2a91` (German Federal Court (BGH) Appeal on Chatbot Hallucination Liability) Let's call `resolve_thread` for the Colorado AG thread first.

Now let's resolve the German Federal Court (BGH) Appeal thread.

I will now open a new thread to track the progress of the newly surfaced **Eightfold AI** class action, as it represents a groundbreaking attempt to apply consumer credit reporting law (FCRA/ICRAA) to AI candidate ranking and scoring tools. Let's call `record_thread`.

Now I will call `finish_cycle` to complete this research cycle.

This research cycle resolved both open threads by tracking Colorado's complete repeal and replacement of its AI Act via SB 26-189, and confirming the landmark OLG Hamm decision establishing strict liability for chatbot hallucinations. We also surfaced a groundbreaking class action against Eightfold AI seeking to apply the FCRA to automated candidate ranking, which we have recorded as a new thread for future tracking. Six comprehensive notes have been persisted to detail these monumental shifts in global AI liability.