Half of proprietary trading firms let their staff use artificial intelligence tools freely, while banks impose limits on employee AI usage, according to a TradingView report. The split marks a sharp divergence in how two segments of the electronic trading industry govern a technology that now touches research, code generation, data analysis and client communication across the front office.

For execution desks, the distinction matters. Proprietary firms, which trade their own capital and answer largely to internal risk committees, can absorb the operational risk of unrestricted AI experimentation. Their traders and quantitative researchers can feed market data into large language models, generate strategy code and automate analytical workflows without a formal approval cycle.

Banks face a different constraint set. As regulated institutions, they must demonstrate control over how customer data, material non-public information and regulatory records flow through third-party AI systems. Unrestricted staff usage creates audit gaps, potential data-leakage paths and recordkeeping exposures under existing supervision rules. Limits on usage are the operational response.

The governance gap has execution-workflow consequences. Buy-side and sell-side counterparties evaluating vendors and liquidity providers increasingly treat AI governance as part of counterparty due diligence. A firm that permits free AI use may move faster on model development and tooling, but it carries a different operational-risk profile than an institution with controlled access.

Compliance timelines frame the divergence. Bank restrictions reflect requirements that already apply — data protection, recordkeeping, model-risk management — extended to a new tool category. Prop firm permissiveness reflects the absence of equivalent external mandates. Neither posture is mandated or prohibited by a specific AI rule at this stage; both represent institutional choices within existing regulatory frameworks.

Technology budgets follow the policy. Firms that allow free usage typically rely on consumer-grade and API-based tools with light internal oversight, a low-cost adoption path. Banks building controlled environments — approved tool lists, data-loss-prevention layers, logging — spend more per seat but generate the audit trails their supervisors expect.

The measured fact from the report is the 50% figure for prop firms permitting free use, set against bank-imposed limits. What remains unmeasured is whether either approach produces better trading outcomes, fewer incidents or lower technology cost per desk. No incident data, firm names or survey methodology accompanied the headline figure.

The open question for market-structure observers is durability. If regulators move from general supervision expectations toward explicit AI usage rules, the permissive half of the prop sector would face the largest adjustment. Until then, the two segments will continue operating under materially different governance regimes for the same underlying technology.