Most companies still treat sustainability claims as a communications asset. From 27 September they are closer to an entry on the risk register, and the numbers behind that shift are not complicated.
That is the date the EU's Empowering Consumers for the Green Transition directive (ECGT, Directive 2024/825) reaches full applicability across all 27 member states. It amends the Unfair Commercial Practices Directive that has governed advertising since 2005, and it moves the burden of proof. Until now a regulator had to show a claim was misleading. From September the advertiser has to show it is substantiated, before publication.
Some claims stop being arguable at all. Generic labels like “eco-friendly”, uncertified sustainability badges and carbon-neutrality claims resting on offsetting go onto a blacklist of practices treated as unfair in all circumstances, with no case-by-case defense. For cross-border infringements, penalties reach at least 4% of annual turnover in the member states concerned.
So the question is not whether your sustainability language is on-message. It is what that language is worth, and what it might cost.
The liability side
The 4% is simple arithmetic. A company with 500 million euros in European turnover is looking at a 20 million ceiling on a cross-border infringement. That is illustrative rather than predictive, since enforcement will vary by member state and by severity, but it sets the order of magnitude.
Now the other side of the ledger: what it costs to know what you have published.
Most large companies cannot produce a current list of their live environmental claims. I have yet to meet one that could. The claims accumulated over years, added by different teams, across websites, product pages, packaging, campaigns, investor decks and press releases, with no owner and no consistent evidence standard applied. Auditing that by hand is a real project, weeks of legal and sustainability time, and it still sits in a different order of magnitude from the fine.
That gap is what makes the decision easy. Even at a low probability of enforcement, not knowing costs more than finding out, and you do not need generous assumptions to get there.
There is a second cost that rarely makes the risk register. A claim withdrawn under pressure is not a quiet event. It gets covered, usually by the trade press, usually at the worst possible moment. The damage lands on the same brand equity the claim was built to create.
The asset side
The more useful half of this is what happens to the companies that get it right.
Until now, a rigorous, evidenced, independently verified claim competed on level terms with a vague one. Both sat in the same ad. The company that ran the life cycle assessment got very little visible credit over the company that wrote a nice sentence.
Enforcement changes that. When unsubstantiated claims carry legal risk, vague language starts leaving the market, and specificity becomes something a competitor cannot copy without doing the work. A claim with evidence behind it stays on the page while others come down.
That is a return, and it goes to whoever moves first. Companies treating September as a compliance chore will come out with thinner claims. The ones treating it as an evidence exercise will come out with fewer claims that are much harder to match.
What a credible program looks like
It is not glamorous. Inventory every live environmental claim across every channel, which is the step that stalls most audits, because the estate is always bigger than anyone expects. Classify each claim by risk, separating the specific and verified from the vague from the outright blacklisted. Map the evidence you actually hold and be honest about the gaps. Then fix, qualify or withdraw. Deleting alone is weak, because a deleted claim tends to reappear in the next campaign; the durable version is a rewrite that says the same thing in language you can defend.
Then gate it. Screen new claims before they publish, or the inventory stays accurate only until marketing ships the next campaign.
AI is genuinely useful for the early steps. Finding claims buried across large volumes of text, sorting them by risk and proposing compliant rewrites are pattern-matching jobs, and pattern matching is what these systems do well. They turn a multi-week audit into something close to real time, which matters a lot when you have weeks.
What they cannot do is judge. Whether the underlying evidence holds, whether a verification is adequate, whether a scope is honest: those stay human decisions, and from September they are legally accountable ones. Nor can a model produce evidence that does not exist. If your recyclability claim has no assessment behind it, run the assessment or drop the claim. Rewording it more carefully is not a fix.
The same tools cut both ways, and that is worth saying plainly. Ask a language model to make a product sound greener and it will hand you fluent, confident, evidence-free copy. That is the practice ECGT exists to stop, only faster. No regulator is going to accept “the tool wrote it”.
Where this leaves you
With the deadline this close, the instinct is to treat the whole thing as a copy exercise. Soften the language, cut the boldest sentences, move on.
I would do the opposite. Treat your published claims as an inventory with a value and a liability attached, count it properly, and put your evidence budget behind the claims worth keeping. That is a defensible position in September and a competitive one afterwards.
Greenwashing was always a problem of scale. The companies that come out of this ahead will be the ones that measured their own exposure before somebody else did it for them.
Enrique Burgos is the founder of Haze, an AI compliance layer for sustainability communications, and a marketing and social-impact advisor (ex-Coca-Cola).
References
Get the latest insights, trends, and innovations to help position yourself at the forefront of sustainable business leadership—delivered straight to your inbox.
Senior Advisor & Co-founder
Haze
Published Sep 17, 2026 9am EDT / 6am PDT / 2pm BST / 3pm CEST