Frontier AI labs dismantled internal ethics oversight this year, creating a void in corporate accountability. Four major laboratories shed ethics personnel and structures designed to evaluate risks before deployment.

The departures reveal a pattern. Labs eliminated dedicated ethics roles, consolidated oversight into existing teams, or reduced the scope of ethics reviews. Internal researchers who departed highlighted a fundamental problem: ethics functions lack enforcement power within organizations focused on speed and market dominance. One departing researcher's observation cuts to the core issue: good intentions cannot substitute for institutional teeth.

These weren't forced exits. Labs restructured voluntarily, framing changes as efficiency improvements or reorganizations. When pressed, companies offered vague assurances that ethics remains "embedded" across their work. That embedding proves hollow without dedicated staff, budget, and authority to halt projects.

The timing matters. As competition intensifies between Anthropic, OpenAI, Google DeepMind, and Meta, ethics functions became seen as friction rather than foundation. Speed accelerates. Ethics slows. In that equation, ethics loses.

The responsibility gap expands. Ethicists can't block decisions. Executives prioritize capability. Boards lack technical depth to challenge course. Regulators lack access to internal processes. External researchers see only published outputs. No single actor holds veto power over deployment.

This structural absence of accountability enables what labs prefer: operational freedom from internal constraints. Ethics becomes rhetoric deployed in policy statements and press releases, absent from actual product decisions. The laboratories reduced ethics not through explicit policy but through attrition and reorganization, creating plausible deniability.

The labs face no immediate consequence for these moves. No regulator has stepped in. No shareholder revolt materialized. The market rewards capability and speed, not caution. Each lab's competitors shrunk their ethics functions too, eliminating any competitive disadvantage to doing so.

This leaves a question without institutional answer: