Snippings

Uber’s approach to this problem is illustrative. They developed a Standardised A2A (Agent-to-Agent) Client on top of the A2A protocol. This client automates the Secure Token Service (STS) JSON Web Tokem ( JWT )exchange and propagation of the actor chain, ensuring the secure path is also the easiest path for developers to implement A2A calls.

xxx

Analogous to the popular term service mesh, the AI Agent Mesh is the data plane where AI agents communicate with each other to complete tasks assigned to them. Within the Agent Mesh and for outbound calls (such as to MCP tools), AI agents rely on JWT tokens minted by the Security Token Service for authentication.

From: Solving the Identity Crisis for AI Agents.

xxx

xxx

The U.S. government earlier this year said it will place any bitcoin it seizes into a strategic reserve. I suppose a crypto reserve sort of makes as much sense as a gold reserve, at least until people lose interest in crypto and/or the power goes off. But there is a problem with the basic idea here: these bitcoins don’t belong to the United States. Assuming that the criminal complaints are proven, then this money is the fruit of crimes committed against millions of victims, and should be returned to them, whether in the form of bitcoin or whatever, not stashed away in Washington.

From: A record Bitcoin haul & crypto comes to the Pitcairn Islands – Coda Story.

xxx

I hadn’t heard the term “shadow AI” before, but I think it’s a very useful addition to the tech lexicon. I came across it while reading about a first-of-its-kind filing by a US bank (CB Financial, the parent of Community Bank) concerning an incident where an employee fed customer data into an unapproved AI application. A law firm called it the first filing to blame a material cybersecurity incident on “shadow AI”, meaning AI tools that employees use without their employer’s approval.

It goes without saying that the use of shadow AI is already widespread and I don’t doubt that people use it all the time. But what do they use it for?

 

xxx

KPMG partner fined £5,000 for using AI to cheat on AI test
A senior member of the Big Four accountancy firm in Australia uploaded training material on artificial intelligence to an AI platform to answer exam questions

From: KPMG partner fined £5,000 for using AI to cheat on AI test.

xxx

xxx

Our data shows that AI use at work is a little messy.

There are gentle tailwinds in the workplace: LLM licenses are bought, training sessions are offered, encouragement from the C-suite is given. But the headwinds are stronger: IT and AI governance; a restricted, limited version of the LLM; reputational risk; fear of AI in general, losing jobs to AI, contravening company policy, looking like you’re cheating.

As a result, “shadow usage” is common. One user reports: “I’m closing tickets 2x faster, my code reviews have fewer issues, and I got praised in my last performance review. But here’s the thing: nobody knows I’m using AI.”

Another wrote: “I built an AI agent to replace myself secretly. Like 50% of the jobs. I used to bring it up to management the way to do it, so the entire org got the benefit. They end up not buying my idea…now I [am] just doing all of these secretly [sic] and spend the spare time on my own side business.”

We see AI mostly being used to achieve modest, uncontroversial wins, some applications in sales and very few where business processes are being fundamentally rethought.

From: How People Are Really Using AI in 2026.

xxx

xxx

Our data shows that AI use at work is a little messy.

There are gentle tailwinds in the workplace: LLM licenses are bought, training sessions are offered, encouragement from the C-suite is given. But the headwinds are stronger: IT and AI governance; a restricted, limited version of the LLM; reputational risk; fear of AI in general, losing jobs to AI, contravening company policy, looking like you’re cheating.

As a result, “shadow usage” is common. One user reports: “I’m closing tickets 2x faster, my code reviews have fewer issues, and I got praised in my last performance review. But here’s the thing: nobody knows I’m using AI.”

Another wrote: “I built an AI agent to replace myself secretly. Like 50% of the jobs. I used to bring it up to management the way to do it, so the entire org got the benefit. They end up not buying my idea…now I [am] just doing all of these secretly [sic] and spend the spare time on my own side business.”

We see AI mostly being used to achieve modest, uncontroversial wins, some applications in sales and very few where business processes are being fundamentally rethought.

From: How People Are Really Using AI in 2026.

xxx

In his book “Making AI Work for Britain”, Alan Brown talks about digital citizenship, going beyond digital literacy and technical knowledge to encompass skills for online safety and privacy management, reasoning about algorithmic bias and the use of tools for civic participation such as citizen juries and open consultation on digital policy. He suggests that experience from Estonia shows the payoff: more than 90% of citizens engage safely in digital government, financial services and health platforms while seeing less fraud, exclusion and mistrust than in the UK. We need to treat digital citizenship as a core public investment rather than a niche educational add-on.

Now, Estonia is Estonia and while it is right to learn lessons from their early success in digital identity (which has nothing to do with blockchains) we are not startng from the same place. We do not have 

I am a long time proponent of Digital Public Infrastructure (DPI). A strong DPI has three foundational components: digital identity, electronic payments and data exchange. Implemented correctly, these are the basis of a safer, more secure and growing economy and I think that implementation should be a priority for any government that takes net welfare seriously.

This is why I was keen to read the The World Bank’s new report “Digital Wallets: A New Paradigm – Convergence of User-Centric Digital Identity, Data Sharing and Payments”, published in May. The report (written by Christopher Tullis, Adam Cooper and David Black) talks very realistically about the benefits of a wallet-centric approach but also about the attendant risks and how to deal with them. In particular, they note how the integration of AI agents into wallets could deliver major benefits by automating routine decisions. AI-powered assistants could also help users understand how their data is being used, for example, by simplifying complex consent forms by translating legal jargon into intuitive language.

Wallets used by AIs are what I would call smart wallets and they may be even more important than payment devices, money managers and digital drivers licence containers. As Kaliya Young and Lucy Yang put it, smart wallets are going to play an important role in an organization’s relationships and interactions with their customers. In common with many other people (eg, Jamie Smith) I would push even further and say that this other channel will become key to business and Jamie’s view of the digital wallet as a safe channel connecting the brand and customers seems a good way to think things.

xxx

This document provides a conceptual and architectural framework for understanding digital wallets, verifiable credentials, and their role in transforming digital identity, data sharing, electronic signing, and digital payments. It is intended primarily for government decision-makers and practitioners seeking to design, regulate, or oversee wallet ecosystems, although many concepts are equally relevant for private-sector participants of these ecosystems – including financial institutions, trust service providers, and technology platforms. This paper focuses on how wallets work, what is new about them, and how they change digital identity, data sharing, electronic signatures, and payments. It also highlights the risks and challenges that implementers must address as these ecosystems grow.

From: Open Knowledge Repository.

xxx

Financial literacy in Britain is at very low levels. In a recent survey, more than half of British adults questioned about Investment Savings Accounts (ISAs) were unable to explain the difference between a cash ISA and a stocks and shares ISA, which seems prettuy fundamental to me. For those aged 25-34 it was seven in 10. I can understand those surveyed when it comes to seemingly endless policy and tax changes that make it near imposisble for normal people to understand what exactly the rules on investing and saving are, but to be honest I cannot see how any kind of financial literacy programme might mitigate such in any imagineable timescale. Given that the survery also showed that a quarter of British adults don’t understand what an interest rate is and that three quarters of them don’t know what an index fund is, the only reaosnable strategy I can imagine is to stop them from managing their money compeltely and get AIs to do it for them instead.

If you think things might improve and that I am being unneccarily dismissive of Brit capabilities when it comes to money, then I must draw your attention to the fact that a third of  Gen Z consumers get their financial advice from TikTok and only a quarter seek advice from actual financial advisors. And how is that working out for them? Not well, it seems. And not only in the UK. Last year investors lost billions of dollars betting (sorry, “investing”) on a handful of US-listed Chinese stocks that plunged in value shortly after being heavily promoted on social media.

I remember when my good friend Ron Shevlin wrote in Forbes that whatever “financial literacy” might be, it seems to do little to improve the financial health of Americans. Same back home in the U.K. It’s time to give up on education and start investing in infrastructure. Instead of financial literacy, let’s get responsible AI into the loop. The idea that I might make a better choice of savings account than event the most rudimentary AI seems fundamentalyl flawed. This doesn’t just apply to savings by the way. Frankly if I never have to talk to a car insurance company ever again it will be too soon:  when it comes to cr, house, health and life insurance, I would cheerfully pay £9.99 per month for a bot to care of them for me in perpetuity.

The big U.S. banks are launching a tokenised deposit network to connect traditional payment rails with digital asset infrastructure. It will be operated by The Clearing House (TCH), which is of course owned by those big banks. The digital asset infrastructure will work through a partnership with an as-yet-undefined vendor with the goal of serving large multinational corporation with use cases including programmable treasury operations, real-time liquidity management and cross-border payments. Banks tend to favor tokenized deposits over stablecoins because they are simply traditional bank deposits represented as digital tokens on the blockchain ,which means that they can provide digital asset payment services within the existing regulatory framework and keep deposits within the banking system.

Banks elsewhere are getting inolved in similar schemes, 

According to the Wall Street Journal, the head of global payments solutions at Bank of America (Mark Monaco) said that customers aren’t necessarily “beating down the door” for tokenized deposits. That may simply be because tokenised deposits are new and bank customers don’t really understand them, or it could be because the bank customers would prefer to see exisiting instant payment networks interconnected.