Application of TEE in AI agents

Source: Dao Blockchain

These days, two AI agents have used a technology that has been around for a while but has never been very popular: TEE.

The full name of TEE is Trusted Execution Environment, trusted execution environment.

This is a hardware-based security solution.

It refers to an execution environment built in a computing device through a combination of hardware and software. This environment is usually used to protect sensitive data and perform critical operations, and it can prevent unauthorized access and malicious operations.

Simply put, if we run software in a reliable TEE environment, it is impossible for outsiders to crack and obtain the data and information generated during the operation.

In the past, this technology was usually used in environments such as IoT devices and cloud computing. In recent years, some encryption projects have also mentioned using this technology, such as wallets.

But overall, I always feel that it is aimed at the enterprise market, and there seems to be no potential for application scenarios for individual consumers, so I basically don’t pay much attention.

The encrypted ecosystem also has a project Phala Network that uses this technology. It provides a cloud computing platform based on TEE technology. In its white paper, it depicts a way to use distributed computing power to provide this execution environment.

I saw this project quite early. I remember that after reading its white paper at that time, I thought of the Filecoin model again, so I think that this so-called way of using distributed computing power to achieve TEE is just another “imaginary” application, purely copying the blockchain.

Recently, two new AI agents, Spore (Spore.fun) and aiPool (@aipool_tee), have surprisingly adopted the technology of this project.

In the previous article, when I introduced the commonly implemented AI agents, I wrote that most of these AI agents need humans to apply for their encrypted wallets. After humans apply for them, they hand over the wallets to the AI agents for use.

In this case, because humans have the wallet private key, humans can completely intervene in the operation of AI agents-------the simplest and rudest way is to directly transfer the assets in the wallet.

So this kind of AI agent cannot be called ‘autonomous’, at least not financially autonomous.

But Spore and aiPool, on the other hand, operate entirely within the TEE environment of the Phala network, with their wallets and keys fully under their control. Humans cannot manipulate their wallets or transfer their assets.

Therefore, in terms of the degree of control over cryptocurrencies, AI agents have now achieved complete autonomy over cryptocurrencies with TEE technology, freeing themselves from human control.

In my opinion, this is really an unexpected surprise of TEE technology in the field of AI agents.

If we continue to develop in this direction, every completely autonomous AI agent will need to run in a TEE environment and generate its own encrypted wallet private key. Then the use cases of TEE technology are far from what I used to imagine, mainly targeting traditional businesses and enterprise users, but will also find huge application space in the quite broad market of AI agents.

As for Phala Network itself, although I still think that the model described in the white paper is far-fetched, it has found its own application scenarios in these two typical cases. This is a remarkable event, which has truly brought the TEE technology into a very promising encryption track.

A few days ago, I was still lamenting in the article: sometimes the development of technology is really unexpected. The usage scenario we yearn for is A, but the real scenario where it truly shines is B.

TEE is also considered a case study.

Speaking of AI agents, it is not an exaggeration to describe its recent development as advancing by leaps and bounds. These developments are not just fantasies in my mind, but they have gradually turned my imagination into vivid reality.