Frontier Labs' Finance Land Grab: Schwab×Anthropic Goes Read-Only as Analysts Give Altruist a 2–5 Year AI Lead
The frontier labs' packaged vertical products (Frontier Labs' Finance Land Grab: Schwab×Anthropic Goes Read-Only as Analysts Give Altruist a 2–5 Year AI Lead through September) hit their first real capability and governance test this week. At Future Proof, Schwab Advisor Services unveiled its exclusive Anthropic deal — "Claude for Financial Advisors," with 18 launch partners including Envestnet, Orion, Addepar, RedTail CRM, and Wealthbox — and was immediately met with a capability critique from the AI-native challenger camp and privacy pushback from RIAs.
The read-only reality. Schwab RIA custody head Jon Beatty told Citywire the deployment is limited to read-only access and is "more of a research function"; letting Anthropic "write" to Schwab systems "would be a phase two" effort, implemented only if advisors "really want" it. Altruist CEO Jason Wenk — who says he turned down the same partnership — was blunt: "Their partnership is read-only, it has no agentic properties, [and] you can basically query your Schwab data, which is something we've been able to do, including on other custodians, for over six months." He added in a Sept 16 LinkedIn post that "What Schwab and Anthropic are building together looks like just skills and connectors so far."
The analyst verdict: years, not quarters. Lex Sokolin (Generative Ventures): "I'd say they (Altruist) are two to five years ahead" — and on legacy custodians catching up: "Measured in calendar time, call it five years or more. Measured in willingness, it could be forever. The tech debt at legacy custodians isn't a backlog waiting for budget. It's load-bearing." Sokolin also flagged the compounding dynamic: "The advantages are cumulative and snowball as we approach technological liftoff. This reprices everything and can create winner-take-all outcomes." Counterpoint from T3's Joel Bruckenstein: "Schwab is in the driver's seat now... It's the gorilla, and they're going to get first crack at taking it to 16,000 advisors" — the exclusive means the RIA ecosystem will build to the Schwab-Anthropic standard.
Data governance is the soft flank. Manish Khatta, CEO of $5B TAMP Potomac: "I guess we are just going to let client data fly into the abyss of Anthropic." Beatty moved to reassure (no account numbers shared), but Nexus Strategy's Tim Welsh framed the structural tension: "'We won't share account numbers with the AI company' pre-answers the question RIAs were already primed to ask. Does the company trying to disintermediate me [with its own expanding RIA presence] now get an AI window into my client base?" This is the same channel-conflict question posed on the data-provider side (see the open thread on premium data providers vs OpenAI's hosted datasets).
The UK regulatory lens arrives. Ben Goss (CEO, Dynamic Planner) mapped the product onto the FCA's Mills Review automation spectrum in Money Marketing: "is this a land grab for Level 5 complete automation at the eventual expense of advisers themselves? Or is it, as billed by Anthropic, a Level 1 or 2 tool – an AI operator or collaborator alongside the human...?" He notes Claude "insists that the human will always have to approve actions (Mills Level 4)" while already "offering more sophisticated ones, such as portfolio re-balancing," and observes the commercial logic: Anthropic wants firms to "spin the meter" in its environment — and gains data that "can help it build better code to automate more," i.e., climb the spectrum. See Regulatory Frameworks and Liability for Agentic Finance: AI Agent Governance, the FCA's Mills Review, and Live Testing Cohorts.
What it means for the landscape: the labs' vertical land grab is now contested on three axes at once — capability depth (read-only connectors vs agentic write access), data governance (who sees RIA client data), and regulatory framing (Mills spectrum positioning). Schwab's exclusivity is explicitly "for now"; the first rival custodian-lab deal remains the trigger to watch.