Jimena Santillán

Jimena SantillánNov 2, 2026

Six stablecoin announcements in seven days. And one thing is becoming clear: the race isn't just about coins. It's about who owns the payments infrastructure. Here's what mattered this week: 1. Visa reported $25B in annualized stablecoin settlement volume, up 10x year-on-year, and 225+ stablecoin-linked card programs. 2. Mastercard + BVNK + Airtm announced wallet-native payments for emerging markets, connecting digital wallets with global payment capabilities. 3. Circle + ACI Worldwide announced a pilot to integrate USDC settlement into Colombia's banking payments infrastructure. 4. Ripple advanced RLUSD through regulatory approval and new custody and distribution partnerships involving BNY, Coinbase and Robinhood. 5. BlackRock + SBI Securities + Startale Group agreed to explore connecting a yen stablecoin to tokenized money-market funds. 6. Tether.io reported nearly $8B in excess reserves and $1.5B in Q3 operating profit. Stablecoins are an economic model, not just technology. My POV? Visa is scaling settlement and cards. Mastercard is connecting wallets to payments. Circle is targeting bank infrastructure. Ripple is building regulated distribution. For banks and PSPs, the biggest question isn't just which stablecoin wins. It's who controls settlement, liquidity, risk and the customer relationship. #Stablecoins #Payments #Banking

Jimena SantillánOct 27, 2026

𝗟𝗮𝘆𝗲𝗿 𝟮 𝗼𝗳 𝗔𝗴𝗲𝗻𝘁𝗶𝗰 𝗣𝗮𝘆𝗺𝗲𝗻𝘁𝘀: 𝗗𝗲𝗹𝗲𝗴𝗮𝘁𝗲𝗱 𝗔𝘂𝘁𝗵𝗼𝗿𝗶𝘁𝘆 & 𝗠𝗮𝗻𝗱𝗮𝘁𝗲𝘀 𝗟𝗮𝘆𝗲𝗿 𝟭 𝗮𝗻𝘀𝘄𝗲𝗿𝗲𝗱: 𝗪𝗵𝗼 𝗶𝘀 𝘁𝗵𝗲 𝗮𝗴𝗲𝗻𝘁? 𝗟𝗮𝘆𝗲𝗿 𝟮 𝗮𝗻𝘀𝘄𝗲𝗿𝘀 𝘁𝗵𝗲 𝗵𝗮𝗿𝗱𝗲𝗿 𝗾𝘂𝗲𝘀𝘁𝗶𝗼𝗻: 𝗪𝗵𝗮𝘁 𝗲𝘅𝗮𝗰𝘁𝗹𝘆 𝗶𝘀 𝘁𝗵𝗮𝘁 𝗮𝗴𝗲𝗻𝘁 𝗮𝗹𝗹𝗼𝘄𝗲𝗱 𝘁𝗼 𝗱𝗼 𝘄𝗶𝘁𝗵 𝗺𝘆 𝗺𝗼𝗻𝗲𝘆? This is where Visa, Mastercard and Amex start to diverge technically. Visa - Once the agent and customer are known, the user authenticates a Payment Instruction. That instruction contains the transaction boundaries, such as merchant, amount and spending restrictions. When the agent later requests a payment credential, Visa Intelligent Commerce checks that request against the authenticated instruction. If it matches, Visa provides an agent specific payment token with network controls attached. At authorization, VisaNet checks again that the actual merchant and amount match the original instruction. So Visa effectively validates the mandate twice: Credential issuance Transaction authorization Mastercard - Connects the mandate more directly to Verifiable Intent. After identity is established, the user’s authorization and instructions are captured in a tamper resistant record linking: 👉 Identity 👉 Intent 👉 Agent action Agentic Tokens can then carry task specific authority, including restrictions by merchant, category, amount, timeframe or usage. For machine commerce, Mastercard goes even further. Businesses define spending limits and execution rules before the agent receives authority to transact. The focus is not only “is this credential valid?” but: Does this action correspond to what the user actually delegated? American Express - Amex has probably the clearest explicit intent object. Its new Intent Intelligence framework creates: ✅ Intent ID ✅Plus ✅Proof of Intent Token The customer first confirms and authenticates the purchase constraints. That Proof of Intent Token is then used to obtain a scoped payment credential. Importantly, the intent stays attached throughout the lifecycle and can support authorization, fraud controls and disputes. And then there is another layer. Platforms such as Sharp Labs can sit above the schemes and decide whether the agent should receive payment authority in the first place. For example: 👀2,000 USD transaction ceiling 👀Approved suppliers only 👀10,000 USD monthly limit 👀Human approval above CHF 1,500 👀No sub delegation 👀Maximum three purchases per day Once that policy says PERMIT, the scheme specific mandate or credential can be created. That distinction matters. The scheme controls the payment instruction. The control plane governs the authority behind it. And that is where delegated authority becomes much bigger than payment authorization. Next: Layer 3, Verifiable Intent. #AgenticCommerce #Payments #Visa #Mastercard #AmericanExpress #AI #Fintech

Jimena SantillánOct 23, 2026

Banks are blocking AI agents. And it's exposing one of the biggest gaps in agentic commerce: TRUST. This week, Reuters and Axios reported that major US banks are restricting third-party AI agents from accessing customer accounts, after detecting agents operating through customer credentials and, in some cases, apparently acting beyond users' instructions. JPMorganChase, Citi, Bank of America, Morgan Stanley and Charles Schwab are among those taking action. The concern isn't simply that AI agents can move money. It's that banks cannot always distinguish between a legitimate customer instruction, an authorized agent and an agent acting outside its mandate. Think about it. If I ask an AI assistant to find me a better savings rate, does that authorize it to transfer €20,000? If an agent has access to my account, who is liable when it makes a payment I never intended? This creates new fraud and risk scenarios: Credential sharing and account takeover: Agents accessing financial accounts through customer credentials. Agent impersonation: Malicious bots pretending to be legitimate AI assistants. Mandate violations: Agents exceeding spending limits, changing beneficiaries or acting beyond consent. Prompt injection and transaction manipulation: Attackers influencing an agent's decisions or payment instructions. Automated abuse: Fraud rings scaling purchases, refunds and account activity through AI. These are emerging risk scenarios, not all confirmed incidents at the banks named above. And an entire fraud and identity landscape is evolving to address them. Visa's Trusted Agent Protocol, Mastercard Agent Pay and Skyfire are developing agent identity, authentication and authorization mechanisms. HUMAN Security, DataDome and Cloudflare are tackling the challenge of distinguishing legitimate agents from malicious automation. Riskified and Forter are extending fraud intelligence into agentic commerce, connecting agents to consumer identities and assessing transaction and post-purchase risk. t54 Labs and Humanos focus on agent-level risk, delegated controls and transaction evidence. Socure, Experian, Prove, Sumsub and Trulioo focus on Know Your Agent and human-to-agent verification. Meanwhile, established players such as Feedzai, Sardine and Sift bring payment fraud detection, behavioral intelligence and automated threat detection capabilities. But here's the distinction I find particularly important: Verifying an agent's identity is not the same as verifying its authority to execute a specific transaction. At Sharp Labs, this is the problem we're working on: a control and risk decision layer for agent-initiated payments. Checking delegated authority, spending limits and transaction context before authorization, with explainable decisions and an evidence trail. The next evolution in fraud prevention won't just be about detecting who is transacting. It will be about proving what they were authorized to do. #AgenticCommerce #Payments #FraudPrevention #Banking #AI