AI Labeling Act of 2026
- Sponsor
- Sen. Schatz, Brian [D-HI]
- Committees
- Commerce, Science, and Transportation Committee (primary)
- Last action
- Jun 24, 2026
Bottom line
The bill establishes a framework for mandatory labeling and detection of AI-generated content and chatbots, aiming to increase transparency and combat misinformation through industry requirements and federal oversight.
What it actually does
This bill mandates that providers of generative artificial intelligence systems label covered AI-generated content with clear, machine-readable disclosures and ensure its detectability. It also requires covered online platforms to identify such content, prevent tampering with disclosures, and combat the 'liar's dividend.' Additionally, it mandates disclosures for AI chatbots and establishes a working group to develop technical standards for AI content transparency.
Proponents argue
Proponents argue that this legislation is essential to combat the growing threat of AI-generated misinformation and deepfakes, ensuring consumers can distinguish between human-created and AI-generated content. They contend it will foster trust in digital information, protect democratic processes, and provide necessary guardrails for the responsible development and deployment of AI.
Opponents contend
Opponents may raise concerns about the technical feasibility and significant compliance costs for AI developers and online platforms, potentially stifling innovation, especially for smaller entities. They might also argue that the definitions and enforcement mechanisms could be overly broad or lead to subjective interpretations, impacting free speech or legitimate creative uses of AI.
The bill is moderately detailed but accessible, allowing an informed reader to grasp its core provisions and implications within a reasonable timeframe.
Section 2(a)
Mandatory Labeling and Machine-Readable Disclosures for AI Content Providers
This provision requires providers of generative artificial intelligence systems to label 'covered AI-generated content' with a clear, conspicuous, and accessible disclosure. It also mandates embedding machine-readable disclosures that identify the AI system, version, creation date/time, and other relevant information, conforming to future standards. Providers must also ensure the content is detectable by widely available tools or provide their own detection tool.
Supporters argue
Supporters argue these requirements are crucial for establishing transparency, combating deepfakes and misinformation, and building public trust in AI. They contend it empowers users to make informed decisions about the authenticity of digital content.
Critics contend
Critics may argue that these requirements are technically challenging, costly to implement, and could stifle innovation, especially for smaller AI developers. Concerns about the precise definition of 'covered AI-generated content' and 'materially alters' could lead to ambiguity or over-labeling.
Tradeoffs
This provision balances the societal need for transparency and combating misinformation against the potential technical and financial burdens on AI developers, and the complexities of creating universal, tamper-proof labeling systems.
Section 2(b)
Online Platform Responsibilities for AI Content
This section requires 'covered online platforms' to clearly and conspicuously identify any AI-generated content displayed on their platforms that incorporates the mandated machine-readable disclosures. Platforms are prohibited from tampering with or removing these disclosures and must offer users the option to make content provenance information readily available. They are also required to make a good faith effort to combat the 'liar's dividend' (false claims that non-AI content is AI-generated).
Section 2(c)
Mandatory Disclosure for AI Chatbots
This provision mandates that any person who makes an artificial intelligence chatbot available to users must include a clear and conspicuous disclosure identifying the system as an AI chatbot.
Section 3(a)
Prohibitions Against Tampering with AI Disclosures
This section prohibits knowingly enabling, facilitating, or concealing the circumvention or falsification of required AI disclosures with the intent or substantial likelihood of deceiving a third party. This includes distributing content without required disclosures or non-AI content with false disclosures, and manufacturing, importing, or offering tools primarily designed or promoted for such circumvention or falsification.
Section 7
Establishment of AI-Generated Content Consumer Transparency Working Group
This section establishes a working group, led by the Director of the National Institute of Standards and Technology (NIST) in coordination with the Federal Trade Commission (FTC), within 90 days of enactment. The group will comprise diverse stakeholders, including federal agencies, AI developers, online platforms, academics, privacy advocates, media organizations, and technical experts. Its duties include developing technical standards for identifying and labeling AI content, ensuring indelible and interoperable disclosures, providing guidance on technical feasibility and detection, and recommending strategies to combat the 'liar's dividend.'
Defines 'covered AI-generated content' as digital content (image, video, or audio) created or substantially modified by a generative AI system such that (A) the system's use materially alters, adds, or removes the meaning or significance a reasonable person would interpret from the content, AND (B) a reasonable person would believe that the content is NOT generated using a generative artificial intelligence system.
Section 8(4)
Why it matters:While placing definitions at the end is standard legislative practice, the specific two-part test for 'covered AI-generated content' is central to the bill's scope and could be easily missed, potentially leading to misinterpretations of what content actually requires labeling.
Case for: Supporters argue this precise definition is necessary to avoid over-labeling mundane or obviously AI-generated content (e.g., simple text generation), focusing instead on content that could genuinely deceive or mislead a reasonable person. It targets the most problematic uses of generative AI.
Case against: Critics might argue that the 'reasonable person' standard is subjective and could lead to inconsistent application or loopholes, making enforcement difficult. The 'materially alters' clause also requires significant interpretation, potentially creating uncertainty for AI developers.
Estimated impact: This definition significantly narrows the scope of content requiring mandatory disclosure, focusing the bill's efforts on potentially deceptive visual, video, or audio content rather than all AI-generated output, thereby reducing the compliance burden for many AI applications.
Exempts covered AI-generated content produced by a provider of a generative artificial intelligence system if the content is generated or used solely for internal research and development purposes and is not intended for public release or commercial deployment.
Section 2(a)(2)
Why it matters:While exemptions are often placed within relevant sections, its concise phrasing might lead a reader to miss its significant implications for AI development practices.
Case for: Supporters argue this exemption is vital to avoid stifling innovation and internal R&D, allowing companies to experiment with AI internally without immediate compliance burdens for non-public content. It prevents unnecessary regulatory friction on early-stage development.
Case against: No strong organized opposition is typically raised against this specific exemption, as it targets non-public content. Some might argue for stricter internal controls or auditing, but the focus on public-facing content is generally accepted as the primary regulatory concern.
Estimated impact: This exemption significantly reduces the compliance burden for AI developers on their internal, non-public work, potentially fostering faster development cycles and experimentation without immediate regulatory overhead.
Defines the term 'liar's dividend' as the benefit a bad actor may receive by falsely claiming that non-AI-generated content is covered AI-generated content. Section 2(b)(4) then requires covered online platforms to make a good faith effort to combat this phenomenon by implementing strategies recommended by the Commission.
Section 8(8) and 2(b)(4)
Why it matters:Definitions are always at the end of a bill, and the mandate is brief. A reader might focus on the positive labeling requirements and miss this crucial defensive measure against a specific form of misinformation.
Case for: Supporters argue that addressing the 'liar's dividend' is crucial to prevent bad actors from discrediting legitimate, human-generated content by falsely attributing it to AI, thereby eroding overall trust in all information, regardless of its origin. It protects the integrity of authentic content.
Case against: Critics might argue that combating the 'liar's dividend' is technically challenging and could lead to platforms inadvertently suppressing legitimate content or facing liability for user claims that are difficult to verify. The 'good faith effort' standard may also be seen as vague.
Estimated impact: Aims to preserve trust in authentic human-generated content by preventing false claims of AI origin, which could otherwise be used to sow doubt and undermine the credibility of reliable information sources.