AI Agents Need IDs: The Identity Arms Race Has Begun

Key Points

• Baselayer raised $35M led by M13 to build identity and fraud infrastructure specifically for autonomous AI agents transacting inside financial systems.
• Palma.ai raised $1.8M pre-seed to centralize permissions governance across AI agents using the Model Context Protocol standard.
• Both companies target the same root problem: legacy identity systems were built for humans and companies, not autonomous software.

Sources

Key Points

• Stripe reports that roughly 70% of its API requests now originate from AI agents, signaling that agentic activity has already crossed the threshold from experiment to mainstream infrastructure load.
• McKinsey projects agentic commerce could redirect $3 trillion to $5 trillion in global retail spending by 2030, making the identity gap a systemic financial risk, not just a product opportunity.
• Visa, Mastercard, American Express, and Shopify have all launched agent commerce protocols in parallel, confirming industry-wide urgency rather than isolated startup speculation.
• The Model Context Protocol, used by Palma.ai as its integration backbone, is an open standard connecting AI models to live business data, meaning the permissions problem is already embedded in enterprise infrastructure at scale.
• Baselayer's claim of preventing over $1 billion in fraud losses across 2,300 institutions gives it rare production-grade credibility in a space where most players are still pre-revenue.

Sources

Sources

winners

• Baselayer's 2,300-institution network gives it a distribution moat that pure-software identity startups cannot replicate quickly, positioning it as a likely acquisition target or category standard-setter.
• Palma.ai benefits from early-mover advantage in MCP governance, a layer that every enterprise deploying Claude, Gemini, or Copilot will eventually need to address.
• Card networks like Visa and Mastercard, which are already building agent commerce protocols, gain legitimacy and urgency for those investments as the identity problem gets formally named and funded.
• Enterprises that adopt agentic governance tooling early will operate with lower fraud exposure and cleaner audit trails, giving them a compliance advantage as regulation catches up.

losers

• Traditional fraud detection vendors built on static rule sets face direct obsolescence pressure as Baselayer explicitly markets against their inability to handle dynamic, agent-driven transactions.
• Enterprises that delay implementing agent identity governance face compounding liability as regulators begin scrutinizing autonomous system activity inside critical infrastructure.
• Mid-market financial institutions without dedicated AI risk teams are the most exposed: they have the agent activity without the oversight tooling.
• Open-source MCP implementations without centralized policy enforcement become security liabilities the moment enterprises move from pilot to production.

implications

• Compliance and legal teams at financial institutions will need to define what constitutes authorized agent action, creating a new category of internal policy work that does not yet have established frameworks.
• Cyber insurance underwriters will likely begin requiring documented agent identity and permissions governance as a condition of coverage, accelerating enterprise adoption of platforms like Palma.ai.
• The FIDO Alliance and x402 Identity Working Group participation by Baselayer alongside Cloudflare, Google, Visa, and Mastercard suggests that standards formation is happening now, and whoever shapes those standards will have lasting structural influence.
• Developer tooling will bifurcate: agent developers will need to build identity attestation into their systems from the start, not retrofit it after deployment.

minority report

• The entire agentic identity category may be solving a problem that hyperscalers solve by default. If Google, Microsoft, and Amazon embed agent identity natively into their cloud stacks, standalone identity infrastructure vendors could find themselves squeezed out before they achieve scale.
• Baselayer's fraud prevention figures are self-reported and unaudited; the $1 billion claim may reflect gross attempted fraud flagged rather than actual losses prevented, which would significantly change its competitive narrative if scrutinized under due diligence.

Sources

second order

• As agent identity standards solidify through bodies like FIDO and x402, early participants including Baselayer, Cloudflare, Google, and Visa gain disproportionate influence over what compliance looks like, effectively writing the rules their own products satisfy.
• The separation of identity from capability from authorization in agentic systems will create an entirely new consulting and integration services market, likely captured first by the large systems integrators already embedded in financial institutions.
• Palma.ai's audit trail feature, which logs every agent action into a tamper-evident record, will become a regulatory baseline expectation rather than a premium differentiator as financial regulators issue guidance on autonomous system accountability.
• Widespread agent identity infrastructure could paradoxically accelerate agentic commerce adoption by removing the trust barrier, pulling forward McKinsey's $3 to $5 trillion estimate by several years.

prediction

• Within 18 months, at least one major data breach or fraud event will be attributed to an unverified AI agent operating inside a financial institution, triggering emergency regulatory guidance and accelerating procurement cycles for companies like Baselayer.
• Baselayer will be acquisition-targeted by a major payments infrastructure company such as FIS, Fiserv, or Mastercard within three years, given its existing distribution across 2,300 institutions and the strategic value of its identity network.
• The Model Context Protocol will become a mandated disclosure item in enterprise AI vendor contracts, elevating Palma.ai's governance layer from optional to contractually required for large enterprise deployments.

minority report

• The agentic commerce boom may stall at the enterprise procurement gate. Large financial institutions historically take five to seven years to integrate new identity infrastructure vendors, meaning the addressable market remains largely theoretical for the next investment cycle regardless of market size projections.
• Regulatory backlash against fully autonomous financial agents could cap the market ceiling entirely. If the EU or US Treasury classifies autonomous transaction execution as requiring human-in-the-loop approval above certain thresholds, the core use case for both companies contracts sharply.

What This Means

Detected Trends

Sources