The Internet has democratized access to information, yet it has only recently scaled in speed and security. In a similar fashion, blockchain is democratizing access to value, yet it is perceived as complex and risky.
Over the course of a few decades, the internet has grown into a global network for accessing information. However, in its early days it used to be a scary place. Over time, a multitude of security protocols have emerged that have made navigating and sharing information safer. Crypto, on the other hand, is still an emerging technology. It already is orders of magnitude safer than in its infancy, but the continuous protocol hacks confirm how fragile the whole stack currently is. One of the recent trends in web3 security hacks has been to leverage social engineering attacks. As onchain protocols become more robust and harder to penetrate, hackers have shifted to exploiting the centralized points of failure of such protocols — i.e., using web2 weaknesses in a project’s own infrastructure.
How does such an attack generally work? Hackers manage to infiltrate a server or a developer’s computer and display a seemingly fine transaction, but with hidden rogue data, so that the developers do not realize they are signing an attack transaction. In cases of concentrated admin control over a protocol, they exploit an intrinsically insecure setup. AI tools also empower the hackers — they only need to spot one bug at any time, while the developer must fix all bugs before deployment.
I see no fundamental difference with how the early internet worked: it was scary and full of exploits until layers of security protocols (HTTPS, secure email standards, firewalls) made it usable for the masses. The same maturation process is happening — and must happen — in blockchain and DeFi.
There is no simple solution to the problem:
Audits are not a silver bullet against bugs in the code
AI audits are becoming more affordable, but they are still experimental
Hackers can impersonate trusted people or websites
One rogue transaction or approval and your assets are gone
This is radically different from the traditional banking world, where transactions are almost manually executed in the background and reversible. However, that comes at a cost: time to transaction finality (often days); slow innovation — boring products; censorship — being expropriated of your assets; access — entire parts of the world population are excluded from financial access; rigidity — each country has different specifications. The onchain world — often referred to as DeFi — gives universal access to financial services in real time across the globe, without national barriers. Currently the legacy financial world is integrating these new technologies into their systems to offer new products to their users, but also to sell their legacy financial products cross-border more easily. IMHO, while you will be able to find a selection of DeFi products in your banking app — thus simplifying your journey and making crypto and blockchain “invisible” — you should see this as further confirmation that the technology is mature to a point where even very risk-averse actors (banks) are actually incorporating it into their own systems. A new gateway to accessing DeFi is AI agents — but this is still very new and, at least in the general setups, intrinsically insecure.
In the current environment, 99.9% of the people I talk to boldly claim that “retail is gone” or “now it’s just institutionals.” I believe they will be proven wrong, and that the trend will be money flowing out of the legacy financial system and running on new rails, like stablecoins and DeFi protocols. Bear in mind that when I refer to DeFi protocols, I am also mentally including Real World Assets (RWAs) — as a protocol does not care whether a token is a cryptocurrency, a utility or security token, or an RWA (in simple terms, an asset whose underlying does not exist onchain and where the token is just a representation of an alien asset). Claude and ChatGPT could offer DeFi access directly, akin to how a search engine gives you direct access to a website. Regardless of what the outcome will be, we all need strong harnesses when using blockchains. And this is only getting more urgent as more people interact with their money through AI agents.
So, can we fix everything? Frankly, I highly doubt we will be able to fix even just 10% of the current weaknesses. The reason? Nobody cares about security until it starts contributing — significantly — to the bottom line. In Ethereumland, the ERC20 token approval feature has been considered by many a bug forever, yet very little has been done to fix that: once you set an approval to an app or a hot wallet and forget about it, it’s there forever — meaning it can easily be exploited if, for example, in 10 years a bug is found in the app or the hot wallet key gets compromised.
Then, are we hopeless? Not quite. In fact, there are a number of emerging tools that are making trading onchain orders of magnitude safer — probably safer than trading in your banking app. Let’s consider a new feature we’ve developed in Rigoblock: the Swap Shield.
The Swap Shield is an open-source API tool — meaning anyone can access it and verify its authenticity by auditing its source code — that offers a critical security feature: protection against fat-finger trades (i.e. protecting a stupid user from his own stupidity). Before any trade is built, it instantly cross-checks the DEX quote against a trusted on-chain oracle price. If the quote is significantly worse than the fair market price (bad liquidity or stale data) or suspiciously better (possible manipulation), the trade is automatically blocked. It works completely separately from normal slippage protection and acts as the first layer in Rigoblock’s two-layer safety stack. I consider myself mostly a stupid user and like to leverage such extra security guarantees — it protects you from fat-finger mistakes, bad liquidity, and even sophisticated quote manipulation.
The Swap Shield is just one of the many security features implemented in Rigoblock to make trading safer. In general, when trading onchain you should assume that everything that could go wrong will go wrong. In the new agentic world, you are practically guaranteed that your agent’s keys — if it owns any keys — will somehow get leaked and compromised (machine security or successful prompt attack).