Temu, Shein, and the end of the old trust model in e-commerce

A few days ago, I came across a discussion about Temu in a Facebook group.

It began with a simple question: is buying toys or children’s clothing there irresponsible? The conversation quickly stopped being about one platform. It moved to certificates, Chinese factories, alleged campaigns against Temu, laboratory tests, trust in regulators, and the familiar argument: “I bought it and nothing happened.”

The longer I read, the more I felt that a much larger issue was hiding underneath.

The problem with Temu is not about Temu.

It is about who should understand risk in global e-commerce, who should control it, and who should be accountable for it.

We do not only buy products. We buy trust

I am not writing from the position of someone who knows better. I do not know whether a particular lunchbox uses a safe plastic. I cannot assess the quality of the pigment used on a toy. I cannot check whether a metal element in a child’s jacket contains nickel, lead, or cadmium. Nor can I independently determine whether the laboratory issuing a certificate actually conducted reliable tests.

Yet I make purchasing decisions every day, just like millions of other people.

This reveals something important: most of us do not buy products alone. We buy trust. We trust that someone has already done their job—the manufacturer, importer, marketplace, retailer, regulator, customs authority, laboratory, or quality-control system.

For years, this model worked reasonably well, at least from the perspective of an average consumer. We entered a shop, took a product from the shelf, and assumed that if it could be legally sold, someone had already checked the basic risks.

Global marketplaces have stretched this model to its limits.

Today, a product can be designed in one country, manufactured in another, sold by a merchant in a third, shipped directly to a consumer in a fourth, while the platform claims it is only an intermediary between the parties.

To the customer, this still looks like an ordinary online purchase. To the accountability system, it is an entirely different reality.

“I bought it and nothing happened” is a very weak safety test

One argument returned particularly often in the Facebook discussion: “I bought it and nothing happened.”

I understand the intuition. This is how we naturally think about safety. If a shirt does not cause a rash, we consider it safe. If a child plays with a toy and nothing happens, the issue appears settled. If a lunchbox does not smell of plastic and does not crack after a week, the problem is difficult to see.

Many risks, however, do not behave like failures. They behave like exposure.

The concern is not limited to spectacular cases in which something explodes, breaks, chokes a child, or causes an immediate allergic reaction. Reports about products bought through marketplaces also mention substances and metals such as phthalates, lead, cadmium, nickel, and chromium. Some are associated with chemical risks, contact allergies, and effects on hormonal or nervous systems.

These are not risks that are easy to see after a single use.

If someone develops a rash the day after wearing a shirt, the causal connection is straightforward. But if a child spends years eating from a container made of untested plastic, wearing clothing from an uncertain source, or playing with toys of unknown composition, nobody will connect a health problem twenty or thirty years later with one childhood purchase.

“I bought it and nothing happened” works well for products that fail immediately. It works far less well for risks that accumulate over time, remain invisible, or depend on dosage, frequency of contact, and individual sensitivity.

A consumer cannot be the toxicologist, importer, and regulator at once

What struck me most in the discussion was how people were trying to settle questions that normally require entire teams of specialists.

Toxicology. Materials chemistry. Certification. Laboratory credibility. Importer accountability. Toy standards. Food-contact safety. The difference between a manufacturer’s declaration and the testing of a specific production batch.

These are complex subjects. Yet we expect an average consumer to make the right decision after seeing a few photographs, reading reviews, checking the price, and perhaps noticing a CE mark in the product description.

That is unrealistic.

The entire purpose of regulation, standards, and control systems is to ensure that we do not all need to become toxicology experts. We did not spend decades building product-safety systems only to tell the consumer at the end: “decide for yourself whether this toy has a safe composition.”

The problem is not that people do not know. The problem is that, at the current scale of trade, they increasingly have to make decisions as if they did.

“It is the same product from the same factory” does not make it the same product

Another popular argument was that most things are manufactured in China anyway.

That is true. A large share of global consumer production takes place in Asia, and many European and American brands use Chinese factories too.

It does not follow that every similar-looking item is the same product.

Two objects may have the same shape, colour, and photograph, and may sometimes even come from the same factory, while differing in plastic, paint, adhesive, dye, metal components, quality control, technical documentation, batch tests, importer responsibility, and complaints procedure.

Most of these differences are invisible to the customer.

In global manufacturing, appearance is not proof that two items are the same product. It may only show that they share a similar design, mould, or photograph.

This is particularly important for children’s goods, food-contact products, and items worn close to the skin. In those categories, invisible differences may matter more than similarities we can see immediately.

The problem is not one product. It is scale

The Polish Consumer Federation commissioned tests of 29 products purchased on Temu and Shein. Its report indicated that some tested clothing, underwear, and accessories contained heavy metals above permitted limits. Discussions of the findings included a figure of 57.6% of products containing dangerous heavy metals within that sample.

This does not mean that half of all products sold on marketplaces are unsafe. That conclusion would go far beyond the evidence. The sample was limited and focused on specific categories.

The findings do show that the risk is not purely theoretical.

Similar signals have appeared elsewhere. European consumer organisations have reported issues involving clothing, jewellery, toys, and food-contact products. In 2025, the Nordic Council of Ministers reported that 71% of 210 online products checked in a chemical-enforcement project did not comply with EU chemical legislation.

The scale of the entire phenomenon is even more important.

The European Commission indicated that approximately 4.6 billion low-value consignments entered the European Union in 2024—around 12 million every day and twice as many as the year before.

At that scale, even a small percentage of problematic goods means an enormous number of real objects in consumers’ homes.

The question is not whether one product from Temu is good or bad. It is whether a system designed for slower, more local, and more controllable trade can handle millions of small parcels every day.

Regulators are beginning to examine systems, not only goods

In 2025, Safety Gate—the EU rapid alert system for dangerous non-food products—recorded 4,671 alerts. The European Commission said this was the highest number in the system’s history.

This can be interpreted optimistically: the system is working, detecting more threats, and warning member states more quickly.

It can also reveal something less comfortable: perhaps we are seeing more clearly than before the limits of a model that seemed sufficient for years.

It is no accident that regulators increasingly look beyond individual products. In May 2026, the European Commission imposed a €200 million fine on Temu for violating the Digital Services Act. The case centred not on one specific item but on insufficient identification, analysis, and assessment of risks related to illegal products on the platform.

This is a significant shift.

A few years ago, the basic question was, “is this product safe?” Today, it is increasingly, “was the entire system designed to limit risk?”

Responsibility moves from an individual listing to platform architecture: seller verification, offer monitoring, product removal, response to reports, data transparency, and cooperation with regulators.

A marketplace is no longer merely a digital noticeboard. It is becoming commercial infrastructure.

Infrastructure cannot pretend that it has no influence over what flows through it.

The counterargument: European shops are not perfect either

The fairest counterargument is that product-safety problems are not exclusive to Temu, Shein, or Chinese platforms.

That is true.

European shops also sell defective goods. Products are recalled, documentation contains errors, manufacturers fail, and suppliers behave dishonestly. A familiar brand or a European retailer cannot provide an absolute guarantee of safety.

This is not a simple division between good Europe and bad China. Such a division would be naive.

The difference lies elsewhere: in accountability, enforceability, and reputational cost.

When I buy from a company operating through a Polish or European entity, I have a better chance of identifying the importer, knowing who is responsible, filing a complaint, understanding who is subject to local inspection, and seeing who may face real legal, financial, or reputational consequences.

That is not a guarantee. It is an additional layer of accountability.

With global marketplaces, that layer often becomes thinner, more fragmented, and harder for an ordinary consumer to enforce.

My personal purchasing filter

As a rule, I do not buy from Temu.

Not because I believe every product sold there is dangerous. That would be too simplistic. I do not want to take on the responsibility of evaluating risks I cannot independently verify.

When someone I know shops on such platforms, I usually suggest a simple filter: choose products with as little impact on everyday life as possible. A garden accessory, decoration, or simple gadget that does not touch food, skin, or a child.

I would not buy anything for children there—no clothing, toys, food-contact products, or items used close to the body.

In these categories, I prefer shops and brands operating through Polish or European entities with real legal and reputational responsibility. Not because I believe they are infallible, but because for products close to the body, food, and children, I do not want price to be the primary optimisation criterion.

This is not a call for panic. It is a personal strategy for reducing risk where I lack the expertise to assess it properly.

Now add AI to the equation

Today, a human still makes the purchasing decision.

They open an app, enter a query, compare photographs, look at the price, read reviews, and sometimes check the seller. This is already difficult, but at least we know who clicked “buy now.”

Tomorrow, an algorithm will increasingly support that decision. The day after, an AI agent may make it.

We will not analyse forty lunchbox listings. We will simply say, “buy a safe lunchbox for my child.” The agent will compare prices, reviews, availability, perhaps certificates, platform rules, and seller history. Then it will choose.

That sounds convenient.

But if it chooses badly, who will be responsible?

The manufacturer? Seller? Marketplace? Model developer? Agent provider? The user who requested the purchase? Or nobody, because everyone is only one link in a long chain?

The question sounds futuristic, but it is the same one that appeared in the Facebook discussion: who is responsible for assessing risk?

AI will not solve this automatically. It may amplify the problem by separating the purchasing decision even further from the human. Today’s consumer can at least see that they are choosing between a cheaper and a more expensive offer. Tomorrow’s agent may optimise according to criteria the user never sees.

Price, delivery time, and star ratings are easy to count. Safety, accountability, and trust are much harder.

Temu is a signal, not an exception

The problem with Temu is not about Temu.

Temu is one of the first mass signals that the product-safety accountability model built over decades is colliding with the reality of global marketplaces.

For years, e-commerce promised greater choice, lower prices, and convenience.

Now a fourth question has arrived: who takes responsibility for risks consumers cannot evaluate themselves?

This will become increasingly important not only for regulators and platforms, but also for brands, retailers, manufacturers, technology providers, and the creators of AI agents.

The future of commerce will not be decided by price, logistics, and personalisation alone.

It will also be decided by trust.

And trust cannot end at the “buy now” button.