🔍Is AI a real threat to DeFi?
valuable weekly insights
GM friends.
Here’s what I’ll cover today:
🔎Is AI a real threat to DeFi?
📊 Crypto chart of the week
🗞️ The latest DeFi news
🔎Is AI a real threat to DeFi?
I promise that I am not going to turn this newsletter into an AI-focused one:)
But I do think it’s important to talk about how AI could impact DeFi, because there’s no denying that AI is getting insanely good at finding code vulnerabilities.
On May 28, Anthropic (the company behind Claude) released the Opus 4.8 AI model.
Just one day later, a security researcher hired by Zcash discovered, using Opus 4.8, a 4-year-old critical bug in Zcash that enabled the minting of an infinite number of $ZEC.
It took him less than 24 hours to find it using AI.
You can imagine how damaging this could have been if the vulnerability had been found first by a hacker trying to exploit it.
Now, bugs inevitably happen.
Even Bitcoin had a few major vulnerabilities in its early days that were eventually patched. But this incident highlighted something important:
AI is quickly reaching the point where it can outperform humans at finding vulnerabilities. According to Anthropic, Mythos, its most powerful advanced AI model, found 10,000 software flaws in its first month.
So it’s easy to see why AI is becoming a major security concern for DeFi.
But on the bright side, I think there are also two positive things about AI that I haven’t seen many people mentioning, and I believe are worth covering:
The public version of Mythos (Fable) that got released a few days ago can’t be used to audit smart contracts by the general public
This is deliberately done by Anthropic to prevent its use in cyberattacks.
Depending on how you see it, this is both good and bad. On one hand, letting anyone use Fable for smart contract audits could help well-intentioned developers find more bugs and patch them. On the other hand, it’s somewhat reassuring that hackers can’t currently use it in order to search for smart contract vulnerabilities.
Some major crypto projects like Ethereum and Solana have reportedly already gained access to the more powerful non-public version of Mythos, which can be used to perform audits and help identify critical bugs
This is obviously great news. If other big projects like Sky, Aave, and Pendle will also eventually gain access to this version of Mythos to perform audits, they could become significantly more secure than they already are.
And I think this is the bull case for DeFi in the AI era.
If the best AI models for security remain gated and are only accessible to vetted teams, auditors, and white-hat researchers, DeFi could eventually reach a point where it becomes incredibly hard for bad actors to find vulnerabilities that haven’t already been discovered and patched by AI.
Now obviously, this depends on what OpenAI and Anthropic will do next.
My hope is that more vetted DeFi teams and auditors will get early access to the most powerful security-focused AI models going forward, allowing them to find and patch vulnerabilities before hackers can exploit them.
If this happens, good actors will gain a major advantage over hackers.
So to sum it up, I think that the DeFi projects that take security seriously will be fine.
As Stani, the founder of Aave, shared, the quality of AI-powered tools for hardening security has improved drastically. The projects using those are much less likely to be exploited via bugs found using AI.
That said, my advice is to be increasingly more careful with the dApps you use.
Stick to DeFi protocols with multiple audits from top security firms, built by teams that give you a feeling that they truly care about security.
The next few years will be a stress test for DeFi, but I think AI will ultimately make it stronger and more resilient.
Chart of the week
Hyperliquid is expected to make $172M in extra annual revenue due to its collaboration with Circle
Crypto meme of the week😂
The latest developments in DeFi
Morpho raised $175M in funding
Curve Finance introduced Llamalend v2 - a new DeFi lending primitive
NEAR Protocol announced confidential perps powered by Hyperliquid
Starknet released STRK20 - a privacy framework for ERC-20 assets
DeFi Saver integrated Hyperliquid and launched S1 of its rewards campaign. Trade perps or open leveraged DeFi strategies to earn a share of its 50,000 USDC rewards pool
Babylon released native Bitcoin-backed borrowing via Aave V4 on testnet
Ondo released Ondo Perps, enabling to trade RWA perps with TradFi liquidity. It offers 24/7 trading access for equities and commodities
Aave proposed a new risk framework following the rsETH exploit
3Jane, a credit-based undercollateralized lending DeFi app, went live to the public
Helius, Mert’s project, announced plans to build a privacy layer for Solana
Polymarket introduced combinatorial positions, which enable to bundle multiple predictions into one bet
Ethena partnered with Centrifuge to add RWAs as part of USDe’s backing
Nado enabled trading perps with tokenized equities as collateral
SBF applied for a presidential pardon
That’s all for this week!
Until next time,
The DeFi Investor
Want to sponsor this newsletter?
Please send me a DM on Twitter (X). I have a sponsorship deck that I can send you.