This article was written by Aleksei Loos, Engineering Manager at Bolt.

Imagine coming into the office, grabbing a coffee, chatting with your colleagues, opening your favourite IDE… and then continuing your day by committing fraud.

Many modern fraudsters don’t fit the old stereotype of a hooded figure in a dark basement. Instead, they operate like IT companies. They have salaries, perks, product managers, QA engineers, overtime policies, and, most importantly, roadmaps. Their job is to find loopholes, exploit systems, and siphon money.

Fraud has become a global business, and its level of organisation is both fascinating and terrifying.

Fighting fraud at Bolt

At Bolt, working against fraudsters feels like a dance. You never quite know their next move, and you can’t plan years ahead on a neat roadmap. There’s no luxury of saying: “We’ll prevent X fraud pattern in Q4 2025 when fraudsters switch to Y technique.” It doesn’t work that way.

Bolt’s Fraud team has taken a risk-based approach. We typically allow the launch of a new feature with a basic set of controls. This gives us protection while not slowing down the business unnecessarily. It also allows us to learn how fraudsters attempt to abuse our products. Because no matter how much research you do upfront, you will always be surprised by their ingenuity.

This can lead to slightly higher fraud losses, as we recently discovered. It turns out that fraudsters had been tracking our updates, and they were the first to take advantage of the new feature before our users did. Fraudsters are usually your early adopters.

Still, taking the calculated risk and not slowing down feature releases is the most optimal strategy.

AI as an accelerant

AI doesn’t just create fraud; it changes the entire game.

Fraudsters can scale faster, automate more, and reach targets with unnerving precision. AI-generated emails, synthetic identities, cloned websites, and data-driven “conversion optimisation” are already the norm. It’s not hard to imagine fraudsters running sophisticated A/B tests on their phishing campaigns.

The technology isn’t new, but AI makes it cheaper, faster, and more scalable.

Fraud-as-a-Service

Fraud has become an industry of its own. It’s no longer just a local gang shaking down shops on the corner. It’s international, and it operates as a service.

Need CAPTCHA solved? There’s a service for that.
Want a fake website to collect credit card numbers? One click, and you get a ready-made package: domain, email server, logos, website content, and even a CDN.
Need stolen credit cards? Buy them in bulk.

Fraud-as-a-Service lowers the barrier to entry. Anyone can plug into this ecosystem, making the problem bigger and harder to contain.

Bolt’s fraud landscape

For Bolt, financial crime is the primary battlefield, encompassing stolen credit cards, money laundering, and payment fraud. We also deal with refund fraud (“I never received my burger”), campaign fraud (abusing discounts with fake accounts), SMS pumping, phishing, DDoS attacks, and scraping of business intelligence. Each category requires its own tools, its own mindset, and its own defences.

Could we stop fraud entirely? Sure, but that would mean adding a lot of hurdles for good users as well and making Bolt harder to use. The challenge is to strike a balance between fraud prevention and a good user experience. In 2025 alone, Bolt’s fraud systems helped prevent losses worth tens of millions of euros across ride-hailing, food delivery, and micromobility, protecting both users and partners.

What needs to change

So, what’s next?

Fraud is likely to worsen before it improves. In the 2000s, antivirus software became the standard protection against malware. Fraud will require its own new generation of defences, and possibly stricter regulation. If every user had to be verified by law, fraud would drop dramatically. But until regulation forces everyone to do it, no company will take the competitive risk. And more data collection also raises privacy risks.

There’s also cooperation. Government-backed CERT (Computer Emergency Response Team) organisations already share information, but we need more collaboration with the private sector. Fighting crime alone is like every house on the block building its own custom alarm system from scratch, while the burglars all share blueprints on Reddit.

And then there’s user friction. We could build walls so high that fraudsters never get through, but neither would real users. If signing up for a scooter takes 15 minutes and five ID checks, people will simply walk to the competitor.

Closing thoughts

Fighting fraud at Bolt means taking ownership of complex problems that have no established playbook and growing rapidly by solving challenges that no one has solved before. The challenge is finding the balance: building defences strong enough to protect people while keeping the experience smooth enough that honest users still want to ride with us. That’s the dance. And it’s one where we want to be in the lead.

Join us!

Help us outsmart fraudsters in the age of AI. Tackle global challenges in data, security, and engineering by building smarter, safer systems that protect millions every day. Explore open roles on our careers page.

Dancing with fraudsters in the Age of AI

We’re hiring!

Recent posts

Progress report: One year of global climate impact