ACAS X Is Already in the Cockpit. The AI-and-ATC Debate Is Three Years Too Late. Part 1.

ACAS X—the Airborne Collision Avoidance System X—has been operating aboard commercial transport aircraft for years, approved by the FAA under Technical Standard Order C-219 and ratified by ICAO in November 2022 as the global next-generation collision avoidance standard. Its decision logic was developed at MIT Lincoln Laboratory using a Partially Observable Markov Decision Process framework, optimized through dynamic programming, with onboard implementations compressed via deep neural networks—machinery that constitutes machine learning applied to safety-critical real-time decisions. When ACAS X issues a Resolution Advisory, the regulatory hierarchy is unambiguous: pilots maneuver contrary to ATC instructions if necessary, because the system resolves conflict geometry at sub-second resolution that a radar display with discrete update intervals cannot match. ICAO Document 9863 codifies this authority structure explicitly. The current media debate over whether AI belongs in aviation safety functions is therefore, as the article's author argues, approximately three years behind a question the federal regulatory apparatus already answered through a disciplined, 15-year development and certification arc.

The contrast the article draws is operationally significant for working pilots and aviation operators. The predecessor system, TCAS II, used hard-coded deterministic rules that produced known deficiencies: nuisance alerts, occasional dissonance with ATC instructions, and limited extensibility to uncrewed aircraft and higher traffic densities. Kochenderfer's team at MIT Lincoln Laboratory reframed the problem probabilistically, accounting for uncertainty in aircraft positions, velocities, and pilot response behavior—precisely the variables that deterministic rule sets cannot adapt to in edge-case encounters. For flight crews operating under Part 121, Part 135, or corporate Part 91 operations, this matters because the RA authority structure has not changed operationally: follow the RA, notify ATC, return to clearance when the conflict is resolved. What has changed is the sophistication of the logic generating that RA, and operators should understand that the neural-network compression implementing ACAS X's lookup tables is not an experimental technology—it is a certified, ICAO-standardized system already aboard their aircraft.

The broader policy context centers on SMART—Strategic Management of Airspace Routing Trajectories—a predictive air traffic management system the FAA is procuring through a three-vendor competition among Palantir Technologies, Thales SA, and Air Space Intelligence, with vendor selection targeted for the end of May 2026 and first operational deployment targeted for September 2026. Transportation Secretary Sean Duffy's public statements that AI will not replace controllers, and the subsequent media coverage characterizing SMART's scheduling-optimization software as a step toward catastrophic autonomy, reflect what the article identifies as a category error: framing the question as human ATC versus autonomous ATC rather than asking which specific functions can be appropriately augmented, at what authority levels, with what failure modes, under what regulatory architecture. For operators and chief pilots monitoring NAS modernization, the SMART deployment timeline is near-term and consequential, particularly for operators whose route structures through congested terminal and en route airspace will be subject to trajectory optimization logic generated by a system not yet in operational service.

The regulatory architecture distinction the article surfaces is the critical analytical thread for aviation professionals evaluating both ACAS X's precedent and SMART's trajectory. ACAS X did not enter the NAS through political reassurance or expedited procurement; it arrived through a development arc that ran from the early 2010s through RTCA standards work, FAA technical order certification, and ICAO ratification before touching a single commercial aircraft. The SMART procurement, by contrast, is being compressed into a competitive vendor selection and a September 2026 initial operational deployment—a timeline that raises legitimate questions about verification depth, failure mode mapping, and controller-interface design that the current public debate, focused on whether AI belongs in aviation at all, is not actually asking. For operators, the practical implication is that the NAS will become more algorithmically managed in the near term regardless of political framing, and understanding the authority levels and failure modes of these systems is not an abstract policy interest—it is operational knowledge that affects how flight departments assess routing decisions, schedule buffer assumptions, and ATC coordination procedures in a changing airspace environment.