Talla Announces Botchain: A Blockchain Security and Compliance Solution For Autonomous Agents

Today I’m excited to announce the private beta program for a new concept — the botchain. As bots proliferate in the enterprise, the problem arises of how to create trust between agent and human, or agent to agent. The botchain allows an autonomous agent to be verified so that bots can be spoofed or spam. It allows an autonomous agent to create a hash of it's current environment (neural model, timestamp, owner, etc) and attach a digital certificate to each task it completes that is immutable on the blockchain. It also allows two autonomous A.I. agents to record transactions between them on an immutable decentralized ledger, and to enforce policies on individual bots to set the limits of their responsibilities. We believe this is a necessity to enable the enterprise bot ecosystem to move forward, and so I have taken some time below to explain why this necessary and what it entails.

Talla is nearly two years old now, and from the very early days, we started talking about how the enterprise bot world would progress. As Talla grew in functionality, many of our customers began asking if we could split the product into multiple bots depending on use case. We quickly realized that if the world starts going this way, there will be a lot of scenarios where enterprise bots need to talk to each other. And since bots are autonomous programs that can learn, make decisions, and will do even more over time - that presents some problems.

The botchain is designed to work on several use cases, on behalf of companies (not consumers) when they need their agents to do something. (We don’t intend to deal with customer to company interactions, but stay B2B focused). The roadmap here is long, and will be detailed in a couple of weeks when we launch our whitepaper, but, here are three things we think are valuable reasons that a botchain should exist.

Audit and Legal Trails — Let’s say I send you an invoice for $500,000. You send it back and say you didn’t buy my service. I say yes, and present you with a contract signed by John Smith at your office. You show it to John Smith, who says he never signed that. I sue you in court for $500,000. How does this get resolved? In most cases, we turn to email archives, which is why email archiving systems exist. Now, as bots start to make simple decisions for us, there will end up being many times when we want to dig in and understand why the decision was made, how the conversation happened, or possibly share the transaction with an auditor, lawyer, or other. While we could build a bot message archiver without a blockchain, we think that since blockchains are essentially a decentralized trust mechanism, that an immutable decentralized ledger of these types of communications is a valuable thing to the ecosystem, and a better way to do it.

On a similar note, what happens when autonomous agents, doing work on behalf of humans, deal with HIPAA data, or have to be part of a SOC2 compliance program - how do you verify what they did?

Bot Identity, Privacy, and Policy — As two autonomous agents transact, how do we know what they are allowed to do? How do we know the bot we are dealing with is really the right bot, and that is has the permission to engage in the transaction it is proposing? A decentralized identity verification mechanism for bots and bot policy makes a lot of sense.

Marketplace For Bot Add-Ons, and Code Enforcement — As artificial intelligence progresses and bots get more complex, there will be opportunities for third parties to release skill and knowledge modules that can upgrade a bot’s brain. We think it dangerous to have one single company control such a marketplace, but given what we have seen in other ecosystems like mobile and social, it is a likely outcome if the marketplace doesn’t live on a blockchain.

Today, APIs deal with program to program communication via clear rules. What happens as they become more intelligent, and the rules are probabilistic? When you deploy a reinforcement learning model to an API endpoint (which is one way to look at an intelligent agent) how do you know what it did, and if it did the right thing?

We will open up a larger beta once the botchain supports use case s #1 and #2. We will also announce the details soon of a token generation event, where the tokens will be a universal unit of access and payment across the botchain. For now, you can keep tabs on the botchain, or get more information at https://talla.com/botchain.

We decided to make this announcement first here on Steemit, because we've bought in to the blockchain, and we think the community here can provide the best feedback to help us shape the future of autonomous agents on the blockchain. We are currently targeting Ethereum and Swirlds for this project, but we welcome your ideas.