If you’ve felt whiplash watching companies promise that artificial intelligence is remaking everything while your day-to-day tools still glitch on basics, you are not alone. A growing chorus of analysts and creators argue that the loudest AI claims often get ahead of the facts. That tension—between glossy marketing and messy reality—sits at the heart of Vanessa Wingårdh’s thesis in her video about companies “lying” about AI.

The point isn’t that AI is fake. It’s that corporate storytelling frequently stretches what a model can do, why a layoff happened, or how a product really works. Unfortunately, regular people pay the price when those stories go unchallenged. Wingårdh’s argument lands because it mirrors something many of us see at work: leadership decks touting “AI-first” futures while teams quietly brute-force solutions the way they always have.

Start with the most basic claim: “We use advanced AI.” The last two years are full of examples where that phrase turned out to be marketing speak. In 2024, the U.S. Securities and Exchange Commission penalized two investment advisers for making false and misleading statements about their use of artificial intelligence.

Regulators called it “AI-washing,” a fancy way of saying the firms advertised capabilities they did not have or exaggerated how much “AI” drove their results. The settlements were modest as far as Wall Street fines go, but the message was loud: say you use AI when you do, say how you use it, and do not invent features for a halo effect. That was one of the first high-profile federal actions to draw a bright line between genuine capability and buzzwords, and it changed how legal and communications teams review AI claims across industries.

Consumer regulators followed a similar path. The Federal Trade Commission has been unusually direct, warning companies to “keep your AI claims in check” and cracking down on deceptive practices where tools are marketed as smarter, safer, or more capable than they are. For marketers and founders, the lesson is simple: if you’re going to claim that your AI delivers specific outcomes, you need substantiation that matches the promise.

If the model only works in narrow use cases, you cannot imply universal magic. What looks like a semantic quibble online becomes a real legal problem in the hands of regulators who police deceptive advertising. That pressure is healthy for buyers and job seekers because it slowly trims the gap between marketing and reality, warns experts like Cooley.

Once you recognize that pressure, Wingårdh’s broader critique comes into focus. Companies often use “AI” to reframe business decisions they would likely have made anyway. A restructuring becomes an “AI transformation.” A roadmap delay becomes “waiting for the next model.” A hiring freeze is justified as “automation efficiency.” The point of calling this out isn’t to deny that AI can change workflows; it’s to make sure the story matches the receipts.

If a firm says AI forced layoffs, you should expect something better than a slide show with futuristic clip art. You should see unit economics, process maps, drastic improvements, and credible before-and-after data. The heart of Wingårdh’s social commentary is that workers and customers deserve better than vague hype cycles, and that the truth is usually visible in the fine print if you know where to look.

A useful way to read AI announcements is to separate three layers: the model, the product, and the process. The model is the statistical engine. The product is the interface that wraps it. The process is how teams actually use it. Most corporate hype lives at the product layer because that’s what screenshots well. Most of the risk and effort live at the process layer because that’s where error handling, audit trails, and human oversight live. This is where laws and standards are pushing companies to be concrete.

For instance, New York City’s local law on automated employment decision tools forces employers to bias-audit AI screening systems if they want to use them for hiring or promotions. That requirement doesn’t ban AI. It asks teams to measure the thing they’re claiming works, to publish the results, and to notify candidates when an automated system is in play. You cannot wave a hand at “smart screening” anymore; you need a testable explanation for what the screening does.

The same shift shows up globally. The European Union’s Artificial Intelligence Act, published in the Official Journal in July 2024, takes a risk-based approach that imposes strict obligations on “high-risk” AI, including many employment and credit uses. It also creates transparency duties for certain limited-risk systems, requiring developers and deployers to tell people they’re interacting with AI in contexts like chatbots and deepfakes. Whether or not you operate in Europe, the ripple effects are obvious: procurement teams and regulators everywhere now have a reference playbook for asking hard questions about model provenance, testing, documentation, and oversight.

For job seekers and employees, that means exaggerated claims will face more scrutiny in the coming years, and real capability will accrue more trust, the EU site warns.

In the United Kingdom, the Competition and Markets Authority has been monitoring foundation models with a focus on competition and consumer protection. Their 2024 updates do not read like breathless press releases. They read like engineers and economists worrying about market power, switching costs, and the risk of misleading claims at scale. Translating that into plain English: if a few players control the infrastructure and everyone else claims “AI-powered” on top without clarity, consumers can be misled and competitors can be boxed out. The CMA’s work adds another layer of skepticism to the hype cycle and reinforces the idea that “we use AI” is not, by itself, a meaningful or defensible statement.

You do not have to be a regulator to see the patterns. Look at how vendors describe accuracy, safety, and bias. When a slide says “near-perfect,” responsible teams will show the test set, the metrics, and the failure modes. When a pitch says “bias-free,” responsible teams will show their procedure for detecting disparate impact, the guardrails in place, and the systems for retesting. This is why bias audits, model cards, and post-market monitoring are more than buzzwords; they are the boring paperwork behind a bold claim.

The EU law even obliges providers of high-risk systems to maintain technical documentation, human-oversight instructions, and monitoring plans so that the performance statements do not turn into a black box that nobody can interrogate. That’s a big cultural shift, and it pulls the center of gravity away from slogans and back toward evidence, ISACA notes.

So where does Wingårdh’s critique land for everyday workers and job seekers? First, treat “AI-first” announcements like any investor would: read the disclosures and look for clarity on what changed. If leadership claims that AI unlocked efficiency, do they explain which processes were automated, how accuracy and quality were validated, and what the humans now do differently? If a hiring team says “our AI is bias-free,” do they link to an independent audit and disclose limitations?

If a product promises “AI-powered insights,” do they define the inputs, the error bounds, and the handoffs between machine and human? In a market where overselling is common, your best defense is to trade in specifics. That mindset helps you evaluate employers honestly and showcase your own credibility without getting caught up in a hype cycle you cannot control, New York City Government adds.

Second, understand that enforcement is catching up. The SEC actions were a narrow slice of finance, but they sent a broader signal that truth-in-advertising rules apply to AI as much as anything else. The FTC’s posture is similar across consumer markets, bluntly reminding companies that there is no “AI exemption” from existing laws. Even the agency’s public comment to the U.S. Copyright Office underscored how quickly AI hype can trip legal wires when companies overpromise on training data, ownership, or performance.

Regulators like the SEC are not trying to kill innovation; they are trying to align promises with proof. That context matters when you see splashy claims in a press release that are not backed by real documentation.

Third, ask yourself who benefits from an AI-framed story. Sometimes the audience is Wall Street. Sometimes it is customers. Sometimes it is employees in a tense all-hands. If the narrative pins layoffs on automation without clear evidence, it is fair to ask whether other drivers—market saturation, interest rates, duplicate teams after acquisitions, or just plain corporate greed—played bigger roles.

Commentary around AI challenges viewers to see beyond “AI did it,” especially when immigration numbers, cost accounting, or product cancellations tell a more complicated story. The point is not to deny that AI can replace tasks. It is to demand that leaders show their work when they claim it replaced jobs. That standard is good for debate, and it is even better for trust.

With that said, remember that real AI progress looks different than slogans. Inside organizations that use models well, the most important wins are often unglamorous: better triage, faster retrieval, cleaner back-office automations, and smarter guardrails. These improvements do not usually “replace whole teams” overnight; they compound into smoother workflows that reallocate human time, which should line up with most company’s claims of increased productivity, not eliminate jobs.

When a company says a tool “eliminated” a category of work, it should be able to show the pre-AI baseline, the post-AI metrics, and the controls that keep outcomes reliable on bad days. The market is starting to reward that maturity and punish the hand-waving. AI will be everywhere, but the kind that lasts is boring on purpose.

Finally, there is a practical takeaway for your career. If you build a reputation for delivering proof instead of buzzwords, you will stand out. On your résumé and LinkedIn, translate “AI” into the specific thing you actually accomplished: the workflow you automated, the error rate you reduced, the minutes you saved per case, the human-in-the-loop checks you designed, and the measurable outcomes for customers.

In interviews, walk through failure modes and how you triaged them. That style reads as competence because it is competence. It also inoculates you against the cynical read of AI claims that Wingårdh criticizes. When everyone else is waving at “intelligence,” you are describing the wiring and the accountability.

If you manage a team, take the same medicine. Map the promises your vendor or internal platform makes to the controls and measurements you own. Publish what you can: documentation, red-team results, bias-audit summaries, and post-deployment monitoring plans.

Tell your stakeholders where the model is strong, where it is brittle, and where humans make the call. The best way to counter the “companies are lying” narrative is to build a culture that does not need to spin. Candor about limitations builds trust faster than another slide deck promising that your chatbot will “revolutionize” the customer journey by Q3.

This conversation will keep evolving as laws tighten and tools improve. The EU’s risk-based regime will roll out in phases. The UK’s competition watchdog will continue prodding dominant players and watching for harms to consumers and smaller rivals. In the U.S., sector regulators and state attorneys general will keep probing claims and training data stories.

So whether you are job hunting, running a team, or choosing vendors, your strategy is the same: favor evidence over adjectives. Ask to see the bias audit if hiring tools are involved. Read the “limitations” section before the demo reel. If a company says “AI forced this change,” request the numbers, not the metaphors.

The future will absolutely include AI; the question is whether it arrives as math you can measure or marketing full of bloated claims you cannot measure. Wingårdh’s message is a nudge to make sure it is the former.

Want help translating your own work into credible, hire-me proof? Grab our free “AI Claims Reality Check” template and our LinkedIn Profile Makeover mini-guide.

References (APA)

European Commission. (2024). AI Act (Regulation (EU) 2024/1689): Regulatory framework for AI. Retrieved from the European Commission’s Shaping Europe’s Digital Future portal. Digital Strategy

European Union. (2024, July 12). The EU Artificial Intelligence Act—Official Journal publication and explorer. artificialintelligenceact.eu. Artificial Intelligence Act

ForHumanity. (2023, October). New York City Bias Audit: An overview and action plan (Local Law 144). forhumanity.center

FTC. (2024, September 25). FTC announces crackdown on deceptive AI claims and schemes [Press release]. Federal Trade Commission

FTC. (2023, March 10). Keep your AI claims in check [Guidance summary, via law-firm analyses]. Cooley; Kelley Drye. cyber/data/privacy insights+1

FTC. (2023, October 30). Comment to the U.S. Copyright Office regarding AI and copyright (P241200). Federal Trade Commission

New York City Department of Consumer and Worker Protection. (n.d.). Automated Employment Decision Tools—Local Law 144 overview. New York City Government

SEC. (2024, March 18). SEC charges two investment advisers with making false and misleading statements about their use of artificial intelligence [Press release No. 2024-36]. SEC

Thomson Reuters. (2024, March 26). AI-washing meets marketing rule, as SEC fines two advisers for false AI claims. Thomson Reuters

Vanessa Wingårdh. (2025). Companies are lying about AI layoffs—Here’s the proof [Video]. YouTube channel page listing. YouTube

Warden AI. (2024, December 10). HR tech compliance: NYC Local Law 144. Warden AI

Securiti. (2024, October 15). NYC Automated Employment Decision Tool (AEDT) law: Overview. Securiti

ISACA. (2024, October 18). Understanding the EU AI Act (white paper): Transparency and post-market monitoring. ISACA

UK Competition and Markets Authority. (2024, April 11). AI foundation models: Update paper. GOV.UK

Cleary Gottlieb. (2024, April 25). CMA publishes update on its initial review into AI foundation models.

Uncategorized