The Federal Trade Commission is reshaping the landscape of artificial intelligence regulation with a series of groundbreaking enforcement actions and policy proposals. While the agency has long policed deceptive marketing, a new frontier has emerged that strikes at the very heart of how AI systems operate: the manipulation of their outputs.
In a move that has sent ripples through Silicon Valley and beyond, the FTC has proposed a policy statement that would deem it a deceptive practice for AI companies to steer their systems’ outputs toward undisclosed ideological or commercial objectives . This is no longer about simply labeling what is and isn’t “AI.” The conversation has shifted to the fundamental question of truthfulness in an age of generative models and agentic systems. This is the latest in a series of aggressive actions that signal a new era of scrutiny, ensuring that the FTC AI news cycle will remain a critical one for businesses and consumers alike.
A Shift in Enforcement Philosophy
For years, the FTC’s approach to regulating technology was largely reactive. It would step in after consumers were harmed, often pursuing companies for making outlandish claims that their products used AI when, in reality, they did not. This practice, known as “AI-washing,” has been a primary focus, but recent developments indicate a much more proactive and, some would argue, expansive interpretation of the agency’s authority under Section 5 of the FTC Act, which prohibits unfair and deceptive acts and practices .
The agency’s current leadership appears to favor an approach that targets conduct and fraud rather than regulating the AI tools themselves . Chairman Andrew Ferguson has described the FTC’s efforts as encouraging “growth in the AI market by targeting bad actors who undermine innovation through deception” . This signals a clear intent to differentiate between beneficial innovation and fraudulent practices, but it also creates a new compliance burden for developers who must now meticulously document their models’ intended behavior and disclose any deviations from a “truthful” baseline.
The Means and Instrumentalities Doctrine Evolves
Central to this evolution is the FTC’s revival of the “means and instrumentalities” (M&I) doctrine. This legal theory allows the agency to hold companies liable for providing the tools that others use to deceive consumers . While this isn’t new, its application to AI tools like large language models represents a significant expansion.
A key test case involved the AI writing assistant Rytr. The FTC alleged that Rytr’s tool, which could generate product reviews, provided users with the means to create fake reviews . This was a controversial case. The FTC later vacated its own consent order, with Chairman Ferguson and others arguing the initial action unduly burdened AI innovation and that the allegations did not amount to deception under the FTC Act . This decision provides a clearer picture of the Commission’s current thinking. The vacated order outlined that the M&I doctrine applies when an entity, for example, provides deceptive marketing material to sellers, who then use it to deceive consumers .
A 2026 case against marketing companies for a fabricated “Active Listening” tool followed this more traditional interpretation, showing the doctrine is still alive and well, but its application must be precise . The FTC is signaling a willingness to pursue the “dealer” who supplies the narrative just as much as the “seller” who propagates it.
New Frontiers in AI-Washing
While the legal theory behind liability is expanding, the agency continues to aggressively pursue classic AI-washing cases. These cases serve as a stark warning to companies that simply slapping “AI” on a product is not a marketing strategy; it’s a liability.
Agentic AI and Unsubstantiated Productivity Claims
A notable example is the FTC’s case against Air AI, filed in August 2025. The agency alleged that the company deceptively marketed its “Odin” conversational AI agent, claiming it could autonomously take actions across thousands of applications and effectively replace human sales representatives . The reality, according to the FTC, was far different. The technology was “faulty” and often couldn’t perform basic functions like scheduling appointments, requiring a massive manual input from users to function at all .
This case is significant for being one of the first consumer protection actions to specifically target deceptive claims regarding “agentic AI” and the potential replacement of human employees . It shows the FTC is keeping pace with AI industry trends, focusing not just on whether a product is AI-powered, but on the substance and verifiability of its performance claims.
The Unending Cycle of Fraudulent Claims
In another case, Growth Cave, the FTC accused the company of misrepresenting its “GrowthBox” software, claiming it would “automate nearly 100% of the process” of setting up an online education course. The reality, again, was that it required substantial manual input . The proposed order in this case bars the defendants from making any misrepresentations that a product “will use artificial intelligence to maximize revenues or otherwise enhance its profitability, effectiveness, or efficiency” .
This boilerplate language is likely to appear in future FTC orders, creating a strict liability standard for any productivity or revenue-generating claims made about an AI product. As one expert noted, companies are being encouraged to ensure they “can substantiate claims made for AI products and that they do not mislead consumers via AI references in marketing materials” .
The Ideological Steering Policy: A Defining Moment
The most significant development in the FTC AI news landscape is the proposed policy statement on AI output accuracy. In July 2026, the FTC published a proposal that would apply Section 5 of the FTC Act to AI companies that “steer” their systems’ outputs “contrary to consumers’ reasonable expectations” . At its core, the policy argues that consumers expect an AI system to provide truthful and accurate outputs. If a company trains or configures its model to prioritize other objectives—such as commercial profit, shaping public opinion, or complying with a state law that mandates embedding certain “equity” or “ideological” values—without clearly and conspicuously disclosing that deviation, it is engaging in deception .
This policy statement is groundbreaking for several reasons. It creates a legal default standard of “truth,” shifting the burden of proof to AI developers to prove they are not manipulating outputs. Under the framework, a company’s motive is irrelevant. Whether the steering is done for profit or to comply with a law like Colorado’s Artificial Intelligence Act, the deception analysis remains the same . This has led to a clear conflict between federal and state regulatory frameworks.
The Federalism Flashpoint
The FTC’s proposal has sparked a fierce debate over federalism. The policy explicitly addresses Colorado’s AI law, which was originally designed to prevent algorithmic discrimination. The FTC argues that a company attempting to comply with such a state law by altering its model outputs to avoid discriminatory outcomes would still be deceiving consumers under federal law . The Commission argues that such state mandates are “impliedly preempted to the extent it conflicts with a federal regulatory scheme” .
President Trump’s Executive Order 14365, “Ensuring a National Policy Framework for Artificial Intelligence,” directed the FTC to take this stance, encouraging a “single, minimally burdensome national scheme over fifty discordant state frameworks” . This has created a tense standoff between the executive branch’s desire for a unified national policy and states’ efforts to regulate AI. Experts believe this political pressure could lead to “a more niche” policy focus on AI, such as deepfakes and children’s safety, where there is more bipartisan agreement .
The Safe Harbor and Compliance Burdens
The policy statement does offer a safe harbor for companies: clear and conspicuous disclosure . However, this safe harbor is not a simple disclaimer buried in the terms of service. To be effective, disclosures must be prominent, persistent, and understandable to the target audience. This has been termed “multi-level communication,” requiring tailored disclosures for technical and lay audiences .
For practitioners, this represents a massive compliance shift. They must now map their model’s design and operational phases against frameworks like the AI Life Cycle Core Principles (AILCCP) to ensure they are meeting their “Truth, Transparency, and Accountability” obligations . Companies are advised to inventory their current output-steering practices, assess whether any steering is motivated by state-law compliance, and then evaluate their disclosure obligations .
Protecting Children: The Companion Chatbot Inquiry
Amidst the focus on output accuracy, the FTC has also turned its attention to a specific, high-risk use case: AI companion chatbots. In September 2025, the agency launched a 6(b) inquiry into seven companies, including Meta, Alphabet, and OpenAI, seeking detailed information on how their products impact children and teens .
The inquiry seeks to understand how these chatbots are monetized, how they generate outputs, and, crucially, what steps companies are taking to monitor for negative impacts like inappropriate content or excessive engagement . Chairman Ferguson stated that “protecting kids online is a top priority,” but also emphasized fostering innovation . This is a delicate balancing act, and the 6(b) study is a precursor to potential future enforcement actions and is informing state-level initiatives like California’s SB 243 . The FTC is positioning itself not just as a consumer protection agency, but as a primary architect of AI safety standards.
Key Takeaways
- Truth is the Default: The FTC’s proposed policy suggests that any AI system steering outputs away from accuracy or objectivity without clear disclosure is deceptive, regardless of motive.
- Vague Claims Are Dangerous: Claims that AI will “maximize revenues” or “replace employees” must be substantiated with hard data.
- State Conflicts are Growing: There is a developing conflict between federal and state AI regulations, especially concerning ideological mandates, requiring careful legal navigation.
- Compliance is Proactive: Companies must document their model steering, implement robust transparency measures, and prepare for increased scrutiny of their marketing language.
- Transparency is the Safe Harbor: The only way to legally steer an AI system is to clearly, conspicuously, and persistently inform consumers of the deviation.
Frequently Asked Questions
What is the FTC’s new policy on AI output accuracy?
The FTC has proposed a policy stating that it is a deceptive practice for AI companies to steer their systems’ outputs toward undisclosed objectives (like profit or ideology) because it violates consumers’ reasonable expectation that the system will provide truthful and accurate results .
What is “AI-washing” and how does the FTC enforce against it?
AI-washing is when companies make false or exaggerated claims about their use of AI to sell products or services. The FTC enforces against this by filing complaints and seeking orders that bar companies from making such misrepresentations. Recent cases have targeted claims about agentic AI and productivity enhancements .
What does the “means and instrumentalities” doctrine mean for AI companies?
It means that a company that provides an AI tool that is used by another party to deceive consumers could also be held liable. While the FTC has pulled back on broad applications of this doctrine, it remains a valid legal theory, particularly when a provider supplies deceptive marketing materials to a seller .
How can a company comply with the FTC’s proposed AI steering policy?
The primary way to comply is through “clear and conspicuous” disclosure. This means prominently and persistently telling users how the AI system is being steered or manipulated, ensuring the information is understandable and not buried in fine print .
What is the FTC’s 6(b) inquiry into AI chatbots?
A 6(b) inquiry is a market study where the FTC demands information from companies without a specific law enforcement purpose. The inquiry into AI companion chatbots seeks to understand their impact on children and teens, examining features like safety practices, monetization, and age restrictions .
Is there a conflict between FTC rules and state AI laws like Colorado’s?
Yes, the FTC has asserted that state laws compelling companies to alter AI outputs to meet certain ideological or equity mandates may be “impliedly preempted” by federal law. The FTC argues that companies cannot use compliance with state law as a defense to a federal deception charge .
Conclusion
The FTC is no longer a passive observer in the AI revolution. Through a combination of aggressive enforcement, novel legal interpretations, and a clear policy framework, it is actively dictating the rules of the road. The central theme is accountability. The days of making grand, unsubstantiated claims about AI capabilities are ending. The new era demands transparency and truthfulness in both marketing and the actual design of AI systems.
Businesses must take note. The “move fast and break things” mentality is a recipe for an FTC investigation. The current regulatory environment requires a robust compliance framework that includes governance, substantiation of marketing claims, and, most importantly, a clear and legally defensible position on how their AI models behave. The agency’s final policy statement, expected later in 2026, will likely set the standard for the industry for years to come, shaping not just American AI policy but potentially influencing global standards. Staying ahead of these developments is not just a matter of legal compliance; it is becoming a prerequisite for earning consumer trust in the age of AI.
You Mey Olso Read
The Great Resurfacing: Why Commercial Epoxy Flooring Is Becoming the Ultimate Sustainable Asset
