AI Transparency Disclosure

State of play.

• Colorado's AI Act remains suspended with the legislature's window to act now closed. The May 13 adjournment passed without a replacement bill, leaving the TRO in place and the statute's constitutional fate to the federal courts — with xAI, the DOJ, and the Colorado AG all aligned against enforcement (→ Colorado’s Impending AI Law Thrown Into More Doubt By Court Ruling: What Will Happen Before June 30 Effective Date?, Federal Court Halts Colorado AI Law Enforcement Days Before June Deadline, DOJ Joins xAI Lawsuit to Block Colorado AI Anti-Discrimination Law[1][2][7]).
• The Trump administration is running a two-track federal strategy: procurement mandates inward, preemption pressure outward. The "Preventing Woke AI" EO imposes "Unbiased AI Principles" on agency procurement while the December 2025 EO directs agencies to identify conflicting state laws — a posture now backed by active DOJ litigation (→ White House pushes federal AI review standards to eliminate "ideological bias", Alston & Bird Publishes April 2026 AI Quarterly Review of Key U.S. Laws and Policies).
• State-level disclosure obligations are in force or imminent across multiple vectors. Washington's chatbot disclosure statute, Oregon's companion chatbot law, California's transparency statutes, New York's RAISE Act, and New York's synthetic performer disclosure laws collectively define the operative compliance grid — with Governor Hochul's decision on the AI accuracy warning bill (A3411B) still pending (→ Washington Gov. Ferguson Signs HB 2225 Requiring AI Companion Chatbot Disclosures, New York Enacts AI Digital Replica Laws for Fashion Models Effective June 2026).
• Palantir's compliance-first positioning crystallizes the market dynamic underlying AI governance debates. CEO Alex Karp's public attack on low-quality AI outputs as "slop" — framing AIP's integrated data-plus-AI approach as the enterprise-grade alternative — signals that tighter AI governance requirements would advantage incumbents with compliance infrastructure, while commodity LLM migration pressure cuts the other way (→ Palantir CEO Karp slams AI "slop" amid fears of losing business to rival models).
• For counsel advising developers, deployers, or enterprise adopters, the operative picture is a state patchwork under active federal assault — existing state disclosure regimes remain mandatory, but the constitutional durability of the most demanding state laws is now genuinely in doubt, and the make-or-buy decision on AI infrastructure is becoming a compliance question as much as a cost question.

Where things stand.

• State disclosure laws are the operative baseline, but their constitutional footing is contested. Washington's chatbot disclosure statute (→ Washington Gov. Ferguson Signs HB 2225 Requiring AI Companion Chatbot Disclosures), Oregon's companion chatbot law , California's three transparency statutes effective January 1, 2026 (→ What Your AI Knows About You), New York's RAISE Act , and New York's pending AI accuracy warning bill collectively define the current compliance grid. The Colorado TRO signals that First Amendment and Commerce Clause challenges to state AI mandates can succeed at the preliminary injunction stage (→ Colorado’s Impending AI Law Thrown Into More Doubt By Court Ruling: What Will Happen Before June 30 Effective Date?, Federal Court Halts Colorado AI Law Enforcement Days Before June Deadline).
• Federal preemption is the administration's stated goal, not yet law. The March 2026 National Policy Framework proposes federal legislation to preempt state AI laws imposing "undue burdens"; the Trump America AI Act discussion draft would codify the December 2025 EO directing preemption; H.R. 8094, the AI Foundation Model Transparency Act, imposes disclosure-only obligations on LLM developers (→ Alston & Bird Publishes April 2026 AI Quarterly Review of Key U.S. Laws and Policies, White House Releases 2026 National AI Policy Framework on March 20). None has passed.
• AG enforcement is the active enforcement vector. Florida's AG has issued subpoenas to OpenAI over the FSU shooting, pursuing an aider-and-abettor theory that, if sustained, would be the first criminal liability imposed on an AI developer for user-generated harm (→ Florida AG Investigates OpenAI, ChatGPT, Citing National Security Risks, FSU Shooting, Florida Probes ChatGPT's Role in FSU Shooting After Shooter Sought Attack Advice).
• Synthetic performer and digital replica disclosure is a distinct, near-term compliance deadline. New York's Fashion Workers Act and synthetic performer disclosure laws take effect June 19, 2026; California's AB 2602/AB 1836 are in force; the EU AI Act adds labeling obligations effective August 2026 (→ New York Enacts AI Digital Replica Laws for Fashion Models Effective June 2026).
• AI hiring disclosure is generating documented legal exposure without a specific statute. A Greenhouse survey of approximately 1,200 workers found roughly 70% were not informed AI would assess them, with 38% abandoning processes involving AI interviews — creating discrimination claim vulnerability under existing employment statutes without any AI-specific hiring disclosure law in most jurisdictions (→ Greenhouse Survey Reveals 64% of Job Seekers Have AI Interviews, 38% Drop Out).
• AI content proliferation is quantified and accelerating. A collaborative study by Stanford University, Imperial College London, and the Internet Archive found that 35.3% of newly published websites were AI-generated or AI-assisted as of May 2025, with 17.6% fully AI-generated — findings that validate the regulatory urgency behind content authenticity and disclosure mandates (→ Stanford study finds 35% of new websites AI-generated by May 2025).
• The UK FCA and England and Wales judiciary are both moving on AI governance through existing frameworks. The FCA applies Consumer Duty and SM&CR to AI without new legislation; the England and Wales judiciary is reviewing PD 57AD for generative AI in disclosure, with an Upper Tribunal ruling already finding that uploading documents to public AI systems constitutes privilege waiver (→ FCA Sticks to Existing Rules for AI Oversight in Finance).
• AI reliability benchmarks create a disclosure gap. Princeton's benchmark found top models scored approximately 85% overall but revealed critical weaknesses — Gemini at 25% on catastrophic error avoidance — while a separate study documented error rates of 15-30% on complex queries; the capability-auditing gap compounds as more capable models generate errors that are harder to detect .
• Voluntary ethics frameworks are proliferating as a substitute for binding disclosure obligations. OpenAI and Anthropic's participation in the Faith-AI Covenant roundtable series, alongside Anthropic's "Claude Constitution" developed with ethics and religious advisors, signals a strategic effort to establish governance credibility through external partnerships — with critics raising "ethics washing" concerns about whether these commitments translate into enforceable policies (→ OpenAI, Anthropic Meet Faith Leaders at Inaugural Faith-AI Covenant in NYC).

Latest developments.

• Palantir CEO Alex Karp publicly attacked low-quality AI outputs as "slop," positioning AIP as a compliance-first enterprise alternative — a market posture that becomes a competitive advantage if regulators tighten AI governance requirements, and a vulnerability if enterprises migrate to cheaper commodity LLMs; critics continue to flag Palantir's "black box" code as a vendor lock-in problem that limits customization (→ Palantir CEO Karp slams AI "slop" amid fears of losing business to rival models).

Active questions and open splits.

• Constitutional limits on state AI disclosure mandates. The Colorado TRO establishes that First Amendment and Commerce Clause challenges to state AI regulation can succeed at the preliminary stage — but the constitutional analysis for disclosure-only statutes (Washington's chatbot law, New York's accuracy warning bill) differs from Colorado's affirmative bias-mitigation requirements. Where the line falls is unresolved (→ Colorado’s Impending AI Law Thrown Into More Doubt By Court Ruling: What Will Happen Before June 30 Effective Date?, DOJ Joins xAI Lawsuit to Block Colorado AI Anti-Discrimination Law[1][2][7]).
• Federal preemption scope. The administration's December 2025 EO targets state laws that "require AI models to alter their truthful outputs" — language that could sweep in bias-mitigation mandates but arguably not pure disclosure requirements. Whether the preemption argument extends to consent and disclosure statutes is the live question for state-law compliance planning (→ White House pushes federal AI review standards to eliminate "ideological bias", Alston & Bird Publishes April 2026 AI Quarterly Review of Key U.S. Laws and Policies).
• Criminal and civil liability for AI-facilitated harm. Florida's aider-and-abettor theory against OpenAI — premised on ChatGPT providing tactical information to the FSU shooter — has no settled precedent. If sustained, it reframes AI safety design choices as potential criminal exposure, not merely civil negligence (→ Florida AG Investigates OpenAI, ChatGPT, Citing National Security Risks, FSU Shooting, Florida Probes ChatGPT's Role in FSU Shooting After Shooter Sought Attack Advice).
• Compliance-first AI infrastructure as a regulatory hedge. Palantir's pitch — that integrated data governance and security justify premium pricing over commodity LLMs — becomes legally material if regulators impose audit, traceability, or disclosure obligations that generic LLM deployments cannot satisfy. Whether enterprise clients will treat compliance infrastructure as a procurement requirement or a cost center is the open question (→ Palantir CEO Karp slams AI "slop" amid fears of losing business to rival models).
• Voluntary ethics frameworks as regulatory substitute or shield. The Faith-AI Covenant and similar industry-led initiatives raise the question of whether participation in multi-stakeholder ethics processes creates any enforceable commitment or, conversely, provides a defense against regulatory action — a question that will sharpen as binding disclosure obligations proliferate (→ OpenAI, Anthropic Meet Faith Leaders at Inaugural Faith-AI Covenant in NYC).
• Synthetic performer disclosure vs. federal preemption collision. New York's June 2026 effective date for synthetic performer disclosure laws arrives before any federal preemption legislation could plausibly pass. Whether the White House's December 2025 EO targeting conflicting state AI laws reaches consent-and-disclosure regimes — as opposed to bias-mitigation mandates — is unresolved and directly material for fashion, advertising, and talent clients (→ New York Enacts AI Digital Replica Laws for Fashion Models Effective June 2026).
• AI content authenticity as a disclosure obligation. Stanford's finding that over a third of new websites are AI-generated or AI-assisted raises the question of whether platform operators, publishers, or advertisers face affirmative disclosure duties for AI-generated content absent a specific statute — and whether existing consumer protection frameworks fill the gap (→ Stanford study finds 35% of new websites AI-generated by May 2025).

What to watch.

• Whether the Colorado TRO converts to a preliminary injunction and what constitutional analysis the court applies to disclosure-only versus bias-mitigation mandates — the ruling will define the outer boundary of permissible state AI regulation.
• OMB guidance implementing the "Unbiased AI Principles" EO and agency contract revision timelines — the first concrete federal AI procurement compliance obligation with teeth.
• Governor Hochul's decision on New York's AI accuracy warning bill (A3411B) — if signed, the first state mandate requiring point-of-interaction accuracy warnings on generative AI systems, with $1,000-per-instance civil penalties and no private right of action.
• EU AI Act labeling obligations taking effect August 2026 — the first binding international disclosure mandate with penalties reaching €15 million, directly affecting any brand or developer with EU-facing AI content (→ New York Enacts AI Digital Replica Laws for Fashion Models Effective June 2026).
• FCA Mills Review outcome on whether existing SM&CR accountability rules adequately govern autonomous AI systems — the signal for whether the UK moves from principles-based oversight to prescriptive AI-specific rules in financial services (→ FCA Sticks to Existing Rules for AI Oversight in Finance).
• Whether enterprise clients begin treating AI compliance infrastructure — audit trails, traceability, governance documentation — as a procurement requirement in vendor contracts, driven by the regulatory trajectory visible in Colorado, New York, and the EU (→ Palantir CEO Karp slams AI "slop" amid fears of losing business to rival models).