Ethereum Goes Private for AI, Dogecoin Adds DeFi, and Blast Calls It Quits

Highlights

• Ethereum Foundation and Open Anonymity Project launched zkAPI for private AI payments
• DogeOS opened a public testnet bringing smart contracts to Dogecoin
• Petrobras is testing Cardano for sustainable fuel and emissions tracking
• Aztec Labs relaunched zk.money as a private wallet on its Ethereum Layer 2
• Blast will shut down after network operating costs exceeded revenue
• Polymarket and Kalshi are expanding further into equity-linked prediction markets
• NEAR Intents halted services after an exploit estimated at about $3.8 million
• Arbitrum paused new Stylus activations over risks linked to AI-assisted attacks
  

Crypto infrastructure is finding its way into more practical parts of the digital economy. Privacy tools are being applied to AI payments, established networks are gaining new functionality, and blockchain systems are being tested in areas ranging from industrial tracking to event-driven financial markets.
‍

At the same time, the operational side remains difficult. Network economics, cross-chain security and new attack methods continue to test whether promising technology can turn into products that are secure, useful and financially sustainable.

Ethereum Takes Privacy Into AI Payments

The Ethereum Foundation and Open Anonymity Project launched zkAPI on Ethereum mainnet, allowing users to pay for access to AI models without directly linking their identity to the questions they submit. Users prepay in USDC, while zero-knowledge proofs confirm access without exposing the same payment information each time a request is made.
‍

The idea addresses a fairly straightforward problem. Paying for an online service usually creates a connection between the user, the payment method and the activity taking place on the account. With AI, that link can be more sensitive because prompts may contain information about work, finances, research or other subjects a user may not want permanently attached to their identity.
‍

What makes zkAPI interesting is that crypto is not the end product here. Blockchain is being used as an underlying payment and privacy layer for a service that already exists outside the industry, which is a more practical role than creating another standalone financial product.
‍

The challenge will be usability. Privacy technology has often struggled when users are expected to understand unfamiliar tools or complex mechanics, so the real test is whether systems such as zkAPI can make the privacy element almost invisible while still doing the job in the background.

Dogecoin Gets a Smart-Contract Layer

DogeOS launched a public testnet for an EVM-compatible zero-knowledge rollup that settles to Dogecoin. The system is intended to support DeFi, stablecoins, prediction markets and games without requiring major changes to the Dogecoin base layer.
‍

That gives DOGE a path into areas where it has historically had little presence. Dogecoin has largely remained a simple transfer network with a large community and a recognisable asset, rather than a platform built around programmable applications. EVM compatibility could make DogeOS more familiar to developers who already work with Ethereum-style smart contracts.

The harder part starts after the testnet. Crypto already has plenty of networks capable of supporting DeFi, games and stablecoins, so DogeOS will need applications people actually want to use, enough liquidity to support them and a reason for DOGE holders to move beyond basic transfers.

Petrobras Tests Cardano in Fuel Tracking

Brazilian state oil giant Petrobras is testing Cardano in two research projects involving sustainable aviation fuel and renewable diesel. The blockchain would be used to create traceable records and reduce the risk that the same emissions reduction is counted more than once.
‍

This is a very different use of blockchain from most products that attract attention in crypto markets. There is no obvious speculative element and no need to build a token economy around the data. The value, if the system works, would come from giving several parties a consistent record they can verify.
‍

That can matter in supply chains where multiple companies handle the same product or rely on the same environmental information. Once sustainability claims have reporting or financial consequences, disagreements over who recorded what and when can become expensive. A shared ledger could reduce some of that friction, particularly if it makes duplication easier to detect.

But Cardano still has to prove that blockchain adds something conventional databases cannot provide efficiently enough. For enterprise applications like this, the technology is most convincing when it quietly improves coordination and auditability rather than becoming the centre of the story.

Aztec Brings zk.money Back

Aztec Labs has relaunched zk.money three years after shutting down the original service, this time running it on the Aztec Ethereum Layer 2. The wallet is designed to keep balances, payment amounts and counterparties private, while users can receive funds through readable handles such as bob.zk.money.
‍

The combination matters because privacy has often come with a usability cost. Users may understand why they want less financial information exposed, but they are less likely to use a product regularly if sending money requires extra steps, unfamiliar addresses or specialist knowledge. Human-readable handles do not solve every usability problem, but they make the experience feel closer to an ordinary payments product.
‍

There is still a difficult balance to manage. Private transactions are attractive precisely because less information is visible, while financial infrastructure also has to deal with operational controls and accountability. For zk.money, the challenge is therefore not only to make privacy technically strong, but to make it practical enough for ordinary use.

Blast Runs Out of Economic Room

Paradigm-backed Blast announced that it will wind down after concluding that the cost of operating the network exceeds the revenue it generates. The Layer 2 once held more than $2 billion in total value locked, but that figure later fell to roughly $32 million.
‍

The contrast is striking because Blast had already demonstrated that it could attract capital. What it did not prove was that those deposits could translate into durable economic activity once the early attention faded. Development, infrastructure and maintenance still have to be paid for even when liquidity moves elsewhere.
‍

That makes the story relevant beyond Blast itself. Crypto often treats TVL and deposit growth as shorthand for success, but those numbers can be misleading when incentives or short-term demand play a large role. A network can look healthy while money is arriving and still struggle to support itself once those flows reverse.
‍

For smaller Layer 2 networks, this is an uncomfortable reality. Users can move capital quickly and developers have many competing ecosystems to choose from, so retaining activity can be much harder than attracting it. Blast reached scale, but scale alone was not enough to keep the economics working.

Prediction Markets Move Closer to Stocks

Polymarket and Kalshi are moving further into equity-linked prediction markets, with contracts connected to companies such as Tesla, Apple and Nvidia. More than $220 million had already been traded in equity-linked markets on Polymarket, showing that the format is drawing meaningful interest.
‍

The attraction is easy to understand. A prediction market reduces a complex view to a defined outcome, allowing users to express an opinion without building a conventional position through shares, options or another financial instrument. That simplicity is part of what makes the format accessible.
‍

Once those contracts are tied directly to listed companies, however, the line between a prediction market and traditional financial exposure becomes less clear. The structure is different, but the economic motivation can be very similar if users are effectively taking views on earnings, corporate events or other stock-related outcomes. That puts more pressure on liquidity, settlement rules and contract wording as the products move closer to mainstream financial markets.

NEAR Intents Faces a $3.8 Million Exploit

NEAR Intents temporarily halted services after an exploit estimated at around $3.8 million. The project said affected users would be fully compensated, reducing the immediate financial damage for users but leaving the underlying security problem to be addressed.
‍

The incident reflects the trade-off built into cross-chain infrastructure. These systems are designed to make moving value between networks easier, but that convenience depends on several components working correctly together. Every additional layer or connection introduces another place where something can fail.
‍

Compensation matters, but reimbursement is still a response rather than a security control. The more important questions concern how quickly the issue was identified, how effectively it was contained and what changes are made before normal activity resumes.

Arbitrum Confronts AI-Assisted Security Risk

Arbitrum temporarily paused new Stylus activations after researchers identified risks involving AI-assisted attacks. Existing programs continued operating while new activations were stopped as the issue was assessed.

The more interesting part is not simply that another blockchain security concern appeared. AI is beginning to influence both sides of the security process, helping developers review code while also giving attackers faster ways to search for weaknesses.
‍

That changes the operating assumptions for blockchain infrastructure. If vulnerabilities can be found faster, projects may have less time between introducing a flaw and seeing someone attempt to exploit it. Deployment controls, testing procedures and emergency responses therefore become more important alongside the code itself.

Arbitrum's response captures that tension well. Open networks are designed to let developers build and deploy quickly, while security sometimes requires exactly the opposite. As AI improves, projects may have to become better at slowing things down when necessary without making normal development unnecessarily difficult.

Bottom Line

The stronger stories point toward crypto infrastructure being used in narrower and more practical ways. Privacy is being applied to AI access and payments, Dogecoin is gaining a route into programmable applications, blockchain is being tested for industrial record keeping, and prediction markets are moving closer to traditional financial activity. These developments are easier to judge because their value depends on whether the product solves a real problem rather than on whether a token attracts attention.
‍

The weaknesses are just as visible. Blast shows that liquidity can disappear long before network costs do, NEAR Intents shows how cross-chain convenience can increase operational exposure, and Arbitrum shows how AI may raise the speed and complexity of security threats. The projects that hold up over time will need more than interesting technology. They will need sound economics, clear market design and systems that can keep working when something goes wrong.