By Alex C. Christou

European supermarkets are throwing away food, Binance has suffered a record cryptocurrency theft, and artificial intelligence is discovering critical flaws in the software underpinning the internet. The economics of cybersecurity are becoming increasingly difficult to ignore.

On Monday morning, supermarket employees across Europe arrived at work to discover that their freezers had developed an unusual enthusiasm for global warming.

In one Swedish supermarket, temperatures reportedly reached 50°C. This is generally considered undesirable for a machine whose primary responsibility is keeping things frozen.

Similar incidents were reported across Belgium, Ireland, Finland and Estonia, where refrigeration equipment began behaving in ways that would have been difficult to explain to somebody buying frozen peas.

Retailers were forced to remove food from sale, dispose of spoiled products and investigate suspected cyberattacks affecting their refrigeration systems.

The incidents followed a larger disruption in France during October, when around 700 supermarket locations reportedly experienced similar problems.

Danish refrigeration manufacturer Danfoss confirmed that its products were involved in the incidents, with older software potentially leaving some equipment vulnerable.

For decades, supermarkets competed on prices, convenience and the quality of their products. They can now add the possibility of someone remotely turning their freezers into ovens. And somehow, this was only one of the week’s cybersecurity stories.

1 Six billion dollars has left the building

While European supermarkets were disposing of food, cryptocurrency exchange Binance was dealing with a somewhat more expensive problem.

On 29 October, the world’s largest cryptocurrency exchange reported a security breach involving approximately $6.3 billion in stolen digital assets, making it the largest reported cryptocurrency theft in history.

The scale is difficult to comprehend. It is an amount that would ordinarily feature in discussions about government expenditure, major corporate acquisitions or the valuation of entire businesses. In this case, it was the subject of a rather unfortunate cybersecurity announcement. Binance responded by temporarily freezing withdrawals, promising to cover customer losses and working with industry partners to freeze stolen assets.

By Monday, net customer withdrawals had reportedly reached $4.7 billion. The exchange subsequently offered additional financial incentives to smaller market makers in an effort to maintain trading activity.

This is where the story becomes particularly interesting from an economics perspective. 1 / 7Market makers are firms that continuously quote prices at which they are prepared to buy and sell assets.

Their willingness to trade helps ensure that investors can execute transactions without dramatically moving market prices.

For their services, market makers receive compensation, frequently through advantageous trading fees or rebates.

In ordinary circumstances, exchanges compete to attract market makers because greater liquidity tends to improve the trading experience. Following a major security breach, however, liquidity becomes particularly important. Investors may withdraw funds, traders may become less willing to hold positions, and uncertainty can reduce the amount of capital available to absorb transactions.

According to figures reported by Bloomberg from trading technology company Talos, the depth of Binance’s Bitcoin order book within 1% of the market price declined from approximately $45 million to $32 million over the 24 hours following the incident.

That represents a reduction of approximately 29%. In practical terms, the market had less readily available liquidity close to prevailing prices.

Larger orders could therefore become more difficult to execute without affecting prices. The distinction between a security breach and a liquidity crisis can become surprisingly small when investors begin questioning whether their assets are safe. It also reveals something familiar about financial markets.

Despite decades of technological development, increasingly sophisticated trading infrastructure and the introduction of cryptocurrencies, investor confidence continues to behave in remarkably traditional ways.

People become worried about their money, and they attempt to withdraw it. The blockchain has apparently made considerable progress without eliminating this particular feature of human behaviour.

Binance Bitcoin market depth

Figure 1: Binance Bitcoin market depth. Source: Talos figures reported by Bloomberg, 2 November 2026, as quoted in the supplied article chat.

Values rounded as reported. This snapshot does not establish how much of the change was caused directly by the breach. Source note for Figure 1: the original market-depth report was not retrieved for this edition; the two values are reproduced faithfully from the supplied chat.

2 The internet’s plumbing has a problem

The most consequential cybersecurity development may, however, have received considerably less attention than the missing billions.

On 21 October, researchers disclosed two critical vulnerabilities in PHP, a programming language that has supported websites and online applications for decades.

PHP is a useful example of how modern technology depends on infrastructure that rarely attracts public attention.

Most internet users have never knowingly interacted with PHP. They have, however, almost certainly interacted with websites and services that depend on it.

The two vulnerabilities, identified as CVE-2026-109618 and CVE-2026-109619, received critical severity scores of 9.2 out of 10.

Depending on the affected configuration, they could expose systems to unauthorised code execution or disclosure of sensitive information.

What makes these vulnerabilities particularly interesting is how researchers discovered them. Security researchers working with artificial intelligence, including systems developed by OpenAI and Google, contributed to identifying and investigating the weaknesses.

Google reported that Gemini 4 Argon helped with vulnerability discovery, exploit development, detection and patch validation.

Separately, OpenAI described how researchers using GPT-5.6 Sol uncovered a critical PHP flaw, with cybersecurity company Wiz independently identifying and reporting both vulnerabilities. The discoveries were followed by a coordinated effort involving researchers, software maintainers and security providers to develop and distribute fixes.

This represents a significant development in how cybersecurity research can be conducted. Historically, discovering complex vulnerabilities required specialist expertise, time and considerable manual investigation.

Modern AI systems are increasingly able to examine software, identify suspicious code behaviour and assist researchers in understanding conditions under which vulnerabilities become exploitable.

The potential productivity gains are substantial. Unfortunately, the internet has spent several decades accumulating software vulnerabilities, so improving the productivity of people looking for them may have some interesting consequences.

3 The economics of finding problems faster

The central challenge is an emerging imbalance between the speed of discovering vulnerabilities and the speed at which organisations can repair them.

A vulnerability may be identified rapidly through automated analysis. Fixing that vulnerability across thousands of organisations requires considerably more coordination.

Software maintainers must develop a correction. Developers must test it. Businesses must identify affected systems, approve updates and deploy them without disrupting critical operations.

For older systems, the situation can be more complicated. Some equipment depends on software versions that no longer receive official security support.

Replacing or updating these systems may require expensive engineering work or the purchase of entirely new equipment. The PHP incident illustrates this issue particularly well.

Google’s technical disclosure identified affected versions extending back to PHP 5.4. While supported PHP branches received security releases, versions that had reached the end of their official support periods did not receive equivalent new releases from the PHP project.

An organisation operating such software may therefore face difficult choices involving compatibility, expenditure and operational disruption.

There is also the problem of timing.

In the PHP case, security organisations were coordinating protective measures ahead of public disclosure. Subsequent reporting identified attempted exploitation before the scheduled release of the fixes, although those observations did not establish widespread successful compromise.

The result is an uncomfortable possibility: the time between vulnerability discovery and attempted exploitation may become increasingly compressed.

For attackers, AI may lower the cost of finding weaknesses. For defenders, AI may also lower the cost of detecting those weaknesses and developing protective measures.

Which side benefits more remains an open question.

But the economic consequences could be substantial.

4 Why businesses sometimes prefer to remain vulnerable

Cybersecurity presents a familiar problem in economics: businesses must spend money today to prevent uncertain losses tomorrow. Consider a company deciding whether to replace an older computer system.

The replacement has an immediate, measurable cost. It may require specialist staff, new equipment, employee training and temporary disruption.

The benefits of improved cybersecurity are considerably more difficult to quantify. If the company invests successfully and experiences no attack, management may struggle to demonstrate precisely how much money the expenditure saved.

It is difficult to present a quarterly earnings slide celebrating the fact that nothing happened. This creates incentives to postpone cybersecurity investment, particularly when businesses are financially constrained or believe an attack is unlikely.

The problem becomes more serious when the consequences of inadequate security extend beyond the organisation making the decision. Economists describe these effects as externalities.

A negative externality occurs when the actions or decisions of one economic participant impose costs on others that are not fully reflected in the original decision. In cybersecurity, a vulnerable supplier can create risks for customers, business partners and entire supply chains.

A supermarket refrigeration system may be purchased and maintained by one organisation, but a security weakness affecting that system can impose costs on manufacturers, employees, consumers and other businesses.

The private incentive to invest in cybersecurity may therefore be weaker than the overall economic benefit of doing so.

This is particularly concerning where infrastructure is highly interconnected. A small vulnerability in a widely used component can become a significant problem when that component is deployed across hundreds or thousands of locations.

The apparent efficiency of standardising technology can also concentrate risk. Using the same software across many systems reduces costs and simplifies maintenance. It can simultaneously create a common point of failure.

The economic trade-off is becoming increasingly difficult to ignore.

5 Are we automating the next cybersecurity crisis?

Artificial intelligence adds a new dimension to this problem.

As AI systems become better at understanding software, they may allow researchers to discover vulnerabilities at a speed that traditional development and security processes cannot easily match.

This could substantially strengthen cybersecurity. Automated tools might identify weaknesses before attackers find them, assist developers in creating patches and help organisations prioritise the systems most urgently requiring attention.

However, these capabilities are unlikely to remain exclusive to defenders. Attackers may also use AI to identify vulnerable systems, develop malicious software or automate elements of their operations.

The implications depend on how capable the systems become, how easily they can be accessed and whether protective technologies improve at a comparable pace.

It is important to recognise that the recent supermarket disruptions and Binance attack have not been established as AI-enabled incidents.

The relevance of AI lies in how the technology could change the future economics of cybersecurity, rather than in assuming that every current security failure was caused by artificial intelligence.

The distinction matters. Public discussion of AI already contains a considerable amount of speculation, often delivered with more confidence than supporting evidence.

Cybersecurity is sufficiently complicated without giving artificial intelligence credit for every malfunctioning refrigerator. Nevertheless, the underlying trend deserves attention.

If AI substantially reduces the cost of finding vulnerabilities, businesses may face a growing need to modernise systems, improve monitoring and shorten the time required to deploy security updates.

That would have implications for corporate spending, insurance, regulation and the economics of critical infrastructure.

6 What happens when the technology gets closer to us?

The implications become especially interesting when considering emerging technologies such as brain-computer interfaces.

A brain-computer interface, or BCI, uses signals recorded from the nervous system to control external devices or communicate information.

Researchers are developing systems capable of helping people with paralysis communicate, control computer interfaces and interact with their environments.

Recent advances in machine learning have made it increasingly possible to interpret complex neural signals and translate them into useful outputs.

But these devices introduce unusually demanding questions about reliability, privacy and security.

An ordinary software failure might prevent somebody from accessing an account or interrupt an online service. For a person who depends on a brain-computer interface to communicate, a system failure could interrupt an essential means of interacting with other people.

The potential consequences become even more serious when considering future systems that combine neural recording with stimulation or other forms of physical intervention.

These systems must address conventional cybersecurity challenges alongside problems involving biological signals, hardware reliability and the interpretation of uncertain data.

A BCI may incorrectly classify a user’s intention without being hacked at all.

Adding connectivity, cloud processing and remote software updates introduces further dependencies that must be carefully secured.

This is one reason why the distinction between laboratory performance and dependable real-world operation matters so much in neurotechnology.

A system may achieve impressive accuracy under controlled experimental conditions while facing considerably more difficult circumstances outside the laboratory.

Signals change, devices move, users behave unpredictably and software must continue functioning under imperfect conditions.

Security is therefore part of a wider question about what it means for an intelligent system to be trustworthy.

For technologies that increasingly interact with the human body, that question becomes particularly consequential. A hacked supermarket freezer creates spoiled food and financial losses.

A compromised financial exchange can threaten confidence in billions of dollars of assets. As computing becomes more closely integrated with human health, communication and everyday decision-making, the potential consequences of failure become increasingly personal.

7 The cost of making everything smart

Much of the modern economy has been built around connecting previously independent systems to the internet. Refrigerators communicate with remote management platforms. Factories depend on automated control systems.

Financial markets operate through interconnected software. Medical devices increasingly incorporate digital monitoring and advanced algorithms.

These developments have created genuine economic benefits. Businesses can reduce costs, improve efficiency, identify equipment failures and deliver services that would previously have been impractical.

Yet greater connectivity also increases the number of systems that must be secured. The challenge facing governments and businesses is to ensure that the productivity benefits of digitalisation are accompanied by sufficient investment in resilience.

That may require stronger software-maintenance practices, clearer responsibilities for technology suppliers and better planning for situations in which connected systems fail. It may also require reconsidering how businesses evaluate cybersecurity expenditure.

The absence of a security incident is a valuable economic outcome, even when it is difficult to measure.

And as artificial intelligence accelerates software development and vulnerability research, the ability to respond quickly may become an increasingly important competitive advantage.

The past week has provided several reminders of what happens when digital infrastructure fails.

European supermarkets have been discarding food. A cryptocurrency exchange has been trying to restore confidence after billions of dollars were stolen.

And security researchers have demonstrated how AI can help uncover dangerous weaknesses in software used throughout the internet. None of these developments means that technological progress should stop.

They do suggest that the economic benefits of increasingly intelligent, interconnected systems come with obligations that organisations cannot indefinitely postpone.

After all, we have spent decades making everything smarter. It would be unfortunate if we discovered that our freezers were easier to hack than to keep cold.

Further reading: For a closer examination of the technical and practical challenges facing brain-computer interfaces, including reliability, neural decoding and privacy, see BCI, Decoded.