The Great AI Illusion – When Innovation Crosses The Line Into Misrepresentation

For decades, technology companies sold us “coming soon.” Artificial intelligence has changed the equation completely. Today, billion-dollar valuations, stock prices, venture capital funding, and public policy decisions are increasingly built on demonstrations of technology that may not fully exist yet.

The problem is not that the future is impossible. Much of it probably is possible. The problem is the growing gap between what companies claim or imply is working today and what is actually happening behind the curtain. At some point, the difference between an ambitious prototype and a misleading representation stops being marketing and starts becoming a forensic issue and greenwashing issue, that is, a regulatory problem with potential significant exposure for organizations.

The Demo Economy

Modern technology launches increasingly resemble movie trailers. A robotic assistant folds laundry flawlessly on stage. A self-driving car navigates city streets effortlessly. An AI system appears to reason like a human. A humanoid robot pours wine, walks naturally, and holds a conversation.

Then, weeks later, details emerge that the robot was remotely controlled, the car followed a pre-mapped route, the AI responses were manually curated, multiple failed attempts were edited out, and that humans were quietly performing many tasks behind the scenes.

The issue is not that prototypes require assistance. Every emerging technology goes through staged development. The issue is whether investors, regulators, and consumers are being given the impression that the technology is autonomous, commercially viable, or production-ready when it is simply not. The distinction matters.

The Tesla Problem

One of the most widely discussed examples involved early demonstrations of autonomous driving technology associated with Tesla. Regulators and courts have increasingly examined whether public statements regarding “Full Self-Driving” capabilities overstated the system’s actual autonomous abilities. Investigations focused on whether drivers and investors may have been left with a misleading impression regarding the level of human supervision still required. The concern is not merely technical.

If a product is marketed as “self-driving,” but requires constant human intervention to prevent collisions, regulators begin asking an uncomfortable question: was the technology being described aspirationally, or deceptively?

That question has significant implications:

• securities disclosure;
• consumer protection claims, including class actions;
• misleading advertising;
• insurance coverage disputes;
• product liability claims; and
• civil and criminal negligence where injuries result from self-driving related vehicle accidents.

The Robot That Wasn’t Really a Robot

We are now seeing the same issues in robotics. Several highly publicized demonstrations of humanoid robots have later been criticized after observers discovered elements of teleoperation or remote human assistance during filming or demonstrations.

Again, there is nothing inherently improper about assisted demonstrations during development. Robotics engineers have done this for years. But the forensic/regulatory issue is disclosure.

If audiences are led to believe that the robot is fully autonomous, that the AI independently made decisions and/or that the system can operate commercially at scale, when in reality hidden human operators are performing the critical functions, the demonstration can move from “visionary marketing” into misleading territory.

That distinction becomes critically important when public companies raise capital, valuations are inflated, governments issue grants, and safety representations influence purchasing decisions.

AI Creates the Perfect Environment for Overclaiming

Artificial intelligence is uniquely susceptible to exaggerated claims because most people cannot independently verify what they are seeing. Unlike traditional technology demos, AI systems often produce outputs that certainly look intelligent even when the underlying capability is weak, heavily scripted, or entirely dependent on human intervention.

This creates an “asymmetric knowledge environment.” The company knows how much is automated, how much is staged, how many humans are involved, how often the system fails and whether the demonstration reflects real-world conditions. But the audience does not. The imbalance creates enormous room for selective disclosure.

The Future Problem

Many technology companies defend themselves by arguing that “We never said it works perfectly today. We said this is where the technology is going.” And that is where things become more interesting.

Much of Silicon Valley operates on future-state representations:

• “This will replace drivers”;
• “This robot will transform labour”;
• “AI doctors will outperform physicians”;
• “General AI is near”; and
• “Human-level robotics is imminent”.

The problem is timing. There is a world of difference between: “We believe this may eventually be possible,” and “This system currently possesses these capabilities.”

Investors are bound to struggle to distinguish between the two, particularly when demonstrations are cinematic, polished, and accompanied by carefully worded language designed to imply capability without explicitly guaranteeing it.

Regulators Are Starting to Notice

Governments are beginning to scrutinize AI and robotics claims more aggressively.

United States

The U.S. Securities and Exchange Commission has increasingly focused on AI-related disclosures to determine whether companies are overstating automation, AI integration, commercial readiness and financial benefits tied to artificial intelligence.

It has specifically warned about “AI washing,” where businesses exaggerate the role or sophistication of AI in products and operations. The Federal Trade Commission has similarly warned companies that false or unsubstantiated AI claims may violate consumer protection laws.

Europe

The EU is pushing toward stricter AI governance frameworks under the proposed AI Act, which focuses heavily on transparency, explainability, risk disclosure and accountability.

The EU’s approach increasingly is treating AI relate representations similarly to the way it currently treats financial disclosure: if you make a claim, you may eventually need evidence to support it.

Canada

Canadian regulators are also paying closer attention to AI-related representations, particularly where investment solicitation, automated decision-making, privacy or consumer safety are involved.

The broader general direction is obvious: regulators are moving toward requiring companies to substantiate technological claims rather than simply market visions of the future.

The Forensic Side Nobody Wants to Talk About

When technology companies collapse, investigators are often brought in after the fact to answer questions such as:

• what actually worked?
• what was manually assisted?
• what did management know internally?
• were investors shown incomplete demonstrations?
• did internal testing contradict public statements?
• were failure rates concealed?
• were engineers warning management privately while marketing claimed success publicly?

That evidence trail almost always exists somewhere. It will be found in emails, Slack messages, engineering tickets, source-code repositories, testing logs, deleted presentations, investor decks, and/or internal risk memos.

The reality behind the marketing often becomes painfully visible during litigation. More and more, this type of evidence, unlike perhaps other types of greenwashing, deceit, misrepresentation or fraud, involves a mixture of technical capability, operational limitations, selective editing and carefully crafted language designed to avoid outright falsehood while still creating a misleading impression. This makes these cases exceptionally difficult

The Insurance and Litigation Fallout

The insurance industry is watching this area carefully. If AI systems are marketed beyond their true capabilities:

• product liability claims proliferate;
• cyber coverage may be triggered;
• directors and officers face added and novel exposure; and
• professional negligence claims also increase.

A company claiming “AI-powered fraud detection” that fails catastrophically could eventually face allegations that the technology never worked as represented, that risks were concealed or that limitations were inadequately disclosed.

The legal exposure could be significant, particularly when human safety is involved.

Innovation Needs Optimism. Markets Need Truth.

There is nothing wrong with ambition. Most transformative technologies sounded impossible before they became real. But there is a critical difference between demonstrating a prototype and presenting an illusion of capability that materially exceeds reality.

The problem is not optimism. The problem is representation. At some point, “coming soon” can become market misrepresentation, and deceptive marketing can become negligent misrepresentation or outright fraud. When billions of dollars depend on perception, the temptation to blur that line becomes very real.

The foregoing provides only an overview and does not constitute legal advice. Readers are cautioned against making any decisions based on this material alone. Rather, specific legal advice should be obtained.

© McMillan LLP 2025