The YFQ-44 is one of two platforms the Air Force chose to prototype the Collaborative Combat Aircraft concept. That selection moved the capability from concept to hardware and put a lot of operational pressure on how these machines will behave near crewed fighters. Pilots and line units do not care about marketing terms. They care about predictable behavior, clear authorities, and robust fail safes when an autonomous jet shows up in their formation.

Where the program stands operationally matters to safety. The YFQ-44 program has been through rapid iteration and ground test activity. The contractor has emphasized autonomy-first development which changes the usual flight-test cadence pilots are accustomed to. That approach is promising for rapid capability gain but it raises two practical safety questions every squadron will ask on day one: how does this aircraft avoid hitting people and aircraft, and who is allowed to tell it to shoot or not to shoot.

From an operational pilot perspective, swarm safety breaks down into four nonnegotiable items:

1) Sense and avoid that you can trust. Cooperative swarms can share tracks and intentions, but any real-world ops will expose systems to jamming, spoofing, sensor occlusion, and friendly clutter. The sense-and-avoid stack must be proven in degraded environments and validated against the same EW threats that have already disrupted civilian navigation in recent years. Redundant sensors, independent kinematic solutions, and explicit fail-over behaviors belong in the baseline design, not as afterthoughts.

2) Clear command, control, and termination chains. A CCA in formation is a weapon platform that must obey unambiguous authorities. That means defined pilot interfaces for tasking and aborts, robust authenticated links, and a hardware kill switch or safe recovery mode that is effective even under contested communications. Rules for human override and the time budget to execute them must be written into tactics and tested to the point where pilots and ground crews can demonstrate them on the range.

3) Integration with manned-air tactics and airspace deconfliction. Line pilots will accept wingmen that fly like wingmen. That requires CCA flight profiles to be predictable at formation scale, to provide status uplinks readable at a glance, and to behave to standard formation geometry and radio discipline when the mission goes sideways. Plan for dedicated formation procedures, standardized visual or datalink cues for separation, and combat-orientated lost-link procedures that return the vehicle to a safe state rather than an uncontrolled intercept.

4) Contested-environment certification and training. You would not hand a novice a jet without training and you should not hand a pilot an AI wingman without rigorous, incremental trust-building events. That means incremented flight envelopes, repeated crossflow drills with manned fighters in instrument meteorological conditions and in EW-rich training areas, and documented acceptance criteria for squadron-level employment. The Air Force and industry are already treating these as test priorities; operational units should insist the same standards apply before tasking CCAs on missions that put crews at risk.

What I see missing from many program-level conversations is the checklist mentality pilots rely on. Translate autonomy behavior into simple procedural items a pilot can call out in the heat of battle. Examples: “CCA lost link, commander designate, engage abort mode alpha;” or “CCA target authorization pending, maintain visual separation, weapons hold.” These are not bureaucratic annoyances. They are the difference between a controllable emergency and a rapidly escalating incident.

Cyber and electronic warfare threats are not theoretical. Modern conflicts have produced tangible examples of GNSS and datalink disruption that affected navigation and unmanned operations. Swarm safety must assume disruption and design predictable degraded modes that prioritize safe separation and recoverability over mission persistence. You cannot build a safe swarm that assumes perfect comms.

Manufacturing scale and production tempo matter to safety too. If the Air Force fields large numbers of CCAs, sustaining consistent software baselines, secure update paths, and a qualified maintenance workforce becomes central to safety risk management. Industry moves to hyperscale manufacturing for these systems will need to bake in configuration control and traceability to avoid a fleet with divergent behaviors. The industrial announcements around scaled production are a sign that volume is coming; program offices should lock down safety and configuration standards before that volume hits the ramp.

Finally, trust remains the social construct that will determine how far commanders will push autonomy. Building that trust takes time and demonstrable results: predictable degraded behavior, transparent decision logging, and evidence from combined manned-unmanned events that crews and commanders can rely on the CCA to act as intended. Without that trust, commanders will not use these systems where they matter most. The Air Force has emphasized this in its public messaging around loyal wingmen.

Operational takeaways for squadron leaders and safety officers

  • Treat early CCA flights as line-level avionics integration events. Document every failure mode and add it to local NATOPS-like procedures. Pilot checklists will be the primary mitigator when things go wrong.
  • Insist on documented, certified lost-link and degraded-sensor behaviors that prioritize safe separation and recovery to a designated recovery area over continuing the mission.
  • Require authenticated, auditable kill and abort mechanisms that function under EW and cyber stress tests. Validate them frequently in rehearsal.
  • Build a training progression that moves from segregated range work to integrated formation flying only when objective safety criteria are met.
  • Push for common interfaces. A single pilot workload display that summarizes CCA state and recommended actions will reduce mistakes. If every vendor uses different, opaque messaging, frontline crews will have to relearn behaviors for each deployment.

Putting autonomous wingmen into the air is an operationally sensible idea. The safety bar, however, has to be set with the same rigor we apply to new engines, weapons and avionics. Predictability, redundancy, clear authorities and a training-first mindset will keep pilots safe and allow commanders to use these systems decisively. As the CCA prototypes move from hangar to runway, the aviation community should focus on those basics first and the flashy swarm demos second.