By Kokab Rahman, Founder & CEO, Radeya Global

The last several days did not produce another model launch or capability demo. They produced operating decisions by large financial institutions and platforms about where third-party AI agents are allowed to stop. On 13 October, JPMorgan Chase CEO Jamie Dimon stated on the bank’s third-quarter earnings call that the firm would not allow “just any outside agent” inside its systems, warning of “rogue agents” and a “liability transfer” for customers who hand over access. A JPMorgan spokesperson confirmed the bank blocks external agents from operating on behalf of retail customers using banking credentials, while still supporting read-only data sharing through authorized open-banking providers such as Plaid.

On or around 16 October, Charles Schwab CEO Rick Wurster told Bloomberg Television that the firm is actively blocking unauthorized AI agents—including Meta’s Muse—from moving client money. He said he receives a report showing these blocks occurring in the “tens and twenties” per week. Schwab will not permit agents to log in with client credentials or use browser automation. Instead, it is directing clients to its own assistant, Charley, and preparing controlled model context protocol (MCP) servers so external models can check balances and perform research without the ability to move money. Wurster framed the issue around control and liability: once a password is shared, “we can’t control it, the client can’t control it, something could go wrong and who’s liable?”

Citigroup has taken a parallel position. The bank stated it would not permit third-party AI agents or assistants to log into a customer’s account or transact directly within its digital banking platforms, while still allowing limited authorized read-only access through certain open-banking providers. Bank of America CEO Brian Moynihan has publicly emphasized the need for a human in the loop to authorize money leaving an account, with the bank working through liability questions for AI-initiated payments.

The same pattern has appeared outside banking. Amazon has blocked Meta’s Muse since mid-September (citing violations of its Conditions of Use and failure to identify as AI) and users have reported blocks on xAI’s Grok Bot for shopping and returns. Delta has stated it has no partnerships allowing third-party AI agents to purchase or book flights. Yelp restricts non-licensed agent access while monetizing data licensing for approved partners. These moves separate agents that can observe or research from agents that can transact, book, or move money.

Anthropic’s equity commitment and the labor file

On 18 October, Anthropic announced it is committing a portion of its shares to a new Anthropic Public Benefit Fund, subject to stockholder approval. The fund starts with 5 percent of shares. The board may add an additional 2 percent for every 50 percent increase in company value above a set baseline. Structured as a wholly owned LLC rather than a traditional nonprofit, the fund is intended for flexibility in large donations to nonprofits or direct distributions. Anthropic explicitly named labor-market impacts from AI as one focus area, building on a prior $200 million commitment to study and address the issue. None of the equity held by the fund will be used for lobbying or political giving. Combined with co-founder pledges and an employee equity-match program, the company stated that roughly 20 percent of shares will be dedicated to public benefit purposes, plus any further amounts tied to valuation milestones. Further details on governance and thresholds are expected in coming months.

Financing stress on the infrastructure side

Separately, CoreWeave’s bonds have traded at distressed levels, with yields rising sharply and credit-default-swap spreads reflecting elevated default concerns. The company carries substantial debt tied to its data-center expansion. The episode has drawn attention to the scale of borrowing supporting AI infrastructure build-outs more broadly, including questions about timing mismatches between capital expenditure and revenue, and the willingness of credit markets to continue financing at prior terms.

What this means for you

These developments sharpen the operating boundary around agents. Large institutions are treating credentialed, autonomous money movement and high-stakes transactions as a control and liability problem rather than a default capability. Read-only or tightly scoped access is being offered in controlled channels; full agent autonomy on customer accounts is not.

• If you are searching, roles that design and operate permission boundaries, audit trails, human-approval gates, and agent-identity management now sit closer to production value. Experience with identity and access systems for non-human actors, open-banking connectors, MCP-style controlled interfaces, or financial-controls design is becoming more directly relevant.
• If you run a team or business, map the external controls before expanding agent reach into systems that hold funds, bookings, or customer relationships. Clarify who authorizes irreversible actions, whether logs are independent of the agent, and how liability is allocated when an outside agent acts under a customer’s credentials.
• Platforms that own customer relationships and monetization (sponsored placements, deposit spreads, booking fees) have clear incentives to prefer their own agents or licensed access over unrestricted third-party ones. Expect more contractual and technical private ordering of this kind.
• Anthropic’s fund structure places a growing equity stake behind efforts that include studying and addressing labor-market effects. For executives and talent teams, this is an additional signal that displacement and transition costs are moving from research papers into concrete resource allocation.
• Confirm details independently. Earnings-call transcripts, company statements, and primary announcements are the clearest sources. Secondary coverage can compress motivations or timelines.

The practical file for both employers and professionals is no longer only what agents can do. It is who is allowed to let them do it, under what constraints, and who remains accountable when the action involves money or a customer relationship.

At Radeya Global we help professionals and businesses turn these market signals into practical advantages—through targeted career strategy, resume and profile optimization, interview preparation, and custom business advisory support. Ready to take action? Explore our career optimization and job search services, or reach out for business consulting support. You can also email us at services@radeya.biz or visit www.radeya.biz.

If an agent could move money or book a flight in your systems tomorrow, which control would you require first—the human approval step, the independent audit log, or the contractual limit on what the agent is authorized to touch?

AIAgents, Banking, Liability, FinancialServices, Anthropic, PublicBenefit, CoreWeave, AIInfrastructure, AgentGovernance, CareerStrategy, BusinessRisk, Technology, Innovation, JobMarket, AutomationControls