Your drone only sees the aircraft that agree to be seen

There is a category of engineering problem where the honest answer is not "this sensor is bad" but "this sensor answers a different question than the one you asked." I spent my Army career as a Counter-IED and Electronic Warfare Officer, which is a job built almost entirely out of that distinction. A receiver tells you about emitters. It does not tell you about the world. If something in the world declines to emit, your receiver is not degraded, it is simply silent, and silence reads exactly like an empty sky.

That is the shape of the problem GAO laid out in GAO-26-107648, published 4 February 2026, a report to the Senate Commerce Committee and the House Transportation and Infrastructure Committee on how drones will communicate with and avoid other aircraft. The report is about small drones, meaning under 55 pounds, because that is what Congress told GAO to look at in the FAA Reauthorization Act of 2024, and because that is what flies low.

The requirement, and the best available answer to it

Today a drone operator who wants to fly beyond visual line of sight needs a waiver or exemption from the FAA, and part of what they must show is how the aircraft will detect and avoid other traffic. GAO asked stakeholders across industry, general aviation and research what technology actually does that job. Four answers came back: ADS-B In, cameras, acoustic sensors, and ground-based radar.

One of those won clearly:

Drone stakeholders generally said that ADS-B In is currently the most effective technology for drones to detect and avoid aircraft that are broadcasting their position data.

Read that sentence twice, because the qualifier at the end is the entire report. ADS-B In is genuinely good at what it does. One operator told GAO their drones detect manned aircraft carrying ADS-B Out from a few miles away, with velocity and position attached, which is enough lead time to plan a real avoidance maneuver rather than a panic. Cameras and microphones cannot come close to that. GAO notes that FAA-approved waivers have mostly relied on ADS-B for exactly this reason.

But ADS-B In is a receiver. It is the same architecture as Remote ID, which I wrote about in Report 079: a cooperative beacon is not a sensor, and it produces a picture of the compliant world rather than the real one. The difference is that Remote ID is a security tool, where the adversary's non-compliance is the whole point. Here nobody is an adversary. The pilot of the aircraft your drone cannot see is not hiding. They are flying a perfectly legal aircraft in perfectly legal airspace, and no rule ever told them to announce themselves.

Who is not broadcasting

GAO's Appendix I lists where ADS-B Out equipment is actually required. It is worth reading as a map of where it is not. The requirement covers Class A, B and C airspace, the Mode C Veil out to 30 nautical miles from the busiest airports up to 10,000 feet, the airspace above and within the lateral boundaries of Class B and C areas up to 10,000 feet, Class E at and above 10,000 feet MSL while explicitly excluding the airspace at and below 2,500 feet above the surface, and Class E at and above 3,000 feet over the Gulf of Mexico out to 12 nautical miles.

Class G, which GAO defines as the uncontrolled airspace that has not been designated as anything else, does not appear on that list at all. Neither does low Class E. Which is to say: the airspace a package-delivery drone spends its whole life in, a few hundred feet over a suburb away from a major airport, is largely airspace where a piloted aircraft has no obligation to broadcast anything.

Then there are the carve-outs inside the requirement itself. Aircraft never certificated with an engine-driven electrical system, which the regulation names as including balloons and gliders, may operate without ADS-B Out in airspace where it would otherwise be required. General aviation groups made the same point to GAO in plainer terms: gliders and ultralights cannot install and operate the equipment because there is no electrical system on board to run it. This is not reluctance. It is physics and airframe certification.

And there is a third category that should interest anyone who works in defense:

Aircraft authorized by FAA that are performing a sensitive government mission for national defense, homeland security, intelligence, or law enforcement purposes are not required to operate ADS-B Out equipment in transmit mode when transmitting would compromise the operations security of the mission or pose a safety risk to the aircraft, crew, or people and property in the air or on the ground.

So the set of aircraft invisible to a drone's best sensor includes a law enforcement helicopter running dark over the same neighborhood the drone is delivering to. When the BVLOS rulemaking committee proposed making non-broadcasting aircraft yield to drones, the FAA pushed back specifically on this, telling GAO the recommendation did not account for low-altitude aircraft that are not ADS-B equipped, or for aircraft on law enforcement or national defense missions authorized to operate with ADS-B Out disabled. The agency's phrasing is careful and worth keeping: the ability of these aircraft to give way "presents a challenge."

Equipage is not the only limit. GAO records that general aviation and research stakeholders, plus a November 2024 DOT research project, found ADS-B In effectiveness reduced in specific conditions: it does not work well at low altitudes where it cannot broadcast around hills, its data are unencrypted (a privacy concern in its own right), and the broadcast frequency can saturate at higher aircraft densities. Line of sight and channel capacity are the two constraints that bite every RF system eventually, and they bite this one lowest and busiest, which is precisely where drone traffic is forecast to grow.

There is even a human reason some pilots do not equip. GAO reports that general aviation groups encourage members to install ADS-B Out, but one stakeholder said pilots worry the FAA or law enforcement could use those broadcasts to monitor violations of noise ordinances. Whatever you think of that concern, it is a real force acting against voluntary equipage, and voluntary equipage is the foundation the whole scheme rests on.

The proposed rule's answer is to invert right-of-way

The FAA's fix is regulatory rather than technical, and it is genuinely clever. In the BVLOS proposal published 7 August 2025 (90 FR 38212, docket FAA-2025-1908, Notice 25-07, a 180-page notice proposing a new 14 CFR Part 108), the agency proposes to adopt part of what its rulemaking committee recommended. The preamble states it plainly:

FAA is proposing to adopt the BVLOS ARC's recommendations related to giving UA right-of-way in shielded areas, giving manned aircraft broadcasting their position using ADS-B Out equipment or electronic conspicuity equipment right-of-way over part 108 UA, and giving part 108 UA right-of-way over manned aircraft who are not broadcasting.

The proposal defines "cooperative" in this context as broadcasting position via ADS-B Out or electronic conspicuity equipment. Under proposed section 108.195 and an amendment to section 91.113, a Part 108 drone would yield to all manned aircraft that are broadcasting, and also to manned aircraft in the densest population category, in Class B or C airspace, or departing from or arriving at an airport or heliport. Proposed 108.195(b) would additionally require the drone to stay a safe distance from any aircraft it must yield to. In shielded areas, which the rulemaking committee framed as within 100 feet of a structure or critical infrastructure, the drone would have right-of-way over all manned aircraft, on the reasoning that manned aircraft are extremely unlikely to be there.

Follow the logic to its end and you get something striking. A piloted aircraft that is not broadcasting would be required to give way to a drone. The pilot cannot see a small drone reliably, and the drone, by the terms of this very report, cannot see the aircraft at all. The rule assigns responsibility to the party less able to discharge it, and it does so knowingly: this is the same objection the FAA itself raised about the committee's version. To its credit the agency is trying to close the gap from the other side, proposing a portable low-cost electronic conspicuity device that a pilot could carry rather than install, usable in aircraft without electrical systems, purely so they can retain right-of-way over drones. The FAA Technical Center tested prototypes in May and June 2024. That is a sensible idea. It is also, note, another voluntary broadcast.

This is the recurring structure of every airspace-control problem I have worked on, and I said the same thing about restricted-airspace designations in Report 107: a rule binds the participant who is listening. It does nothing to the glider with no radio, and nothing to the aircraft told to go quiet.

A second sensor, and what it costs

The obvious answer is ADS-B "Plus": use ADS-B In for cooperative traffic and add a camera, acoustic sensor or ground radar for everything else. General aviation stakeholders told GAO this would be the safest approach, and in a 2023 letter to the FAA they argued all drones should carry technology able to detect both equipped and unequipped aircraft. FAA officials agreed it adds real redundancy.

Then GAO does the thing I wish more technology assessments did, and prices it. Drone stakeholders pointed out that bolting on sensors increases the size and weight of the aircraft, and one made the argument that this can reduce net safety, because a heavier drone is more dangerous in the collision you failed to prevent. That is a genuine safety trade, not a cost complaint, and it does not have an obvious answer.

The added sensors also underperform. Research stakeholders told GAO that acoustic sensors pick up background noise such as wind or highway traffic, which is the same limitation I covered in Report 043 pointed in the opposite direction. Cameras, per one stakeholder, have difficulty differentiating aircraft from birds and are affected by clouds. I wrote Report 085 about radar confusing drones with birds; it is a small pleasure to report that the drone's camera returns the favor. Ground radar has short range and cannot be moved easily, which pins the operator to one area, though a public safety operator noted radar's real advantages: it runs 24 hours and works better at night than cameras.

None of these is a bad sensor. Each has a passband, a failure mode and an environment it hates, and the honest way to read the ADS-B Plus proposal is that it stacks four partial answers and hopes the gaps do not line up.

The jamming case, raised by the FAA itself

One detail in GAO's report deserves more attention than it will get. Explaining why a second technology is worth the weight, FAA officials gave two reasons. The first is ordinary equipment failure on a single aircraft. The second is that if there is a broader problem, such as GPS spoofing or jamming, the other sensors would still be available to detect aircraft.

That is a federal aviation regulator listing GNSS interference as a design case for civil drone collision avoidance, and it is correct. ADS-B Out works by an aircraft broadcasting a position it derived from GPS. Deny or corrupt the GPS and you have not silenced the broadcast, which would at least be obvious. You have left a confident aircraft broadcasting a confident and wrong position, and a drone dutifully avoiding a place where nothing is. That failure mode, where the system keeps reporting through the compromise, is the whole subject of Report 020, and the reason a navigation system needs something that does not depend on a satellite, which I covered in Report 049. A camera or a microphone is not a great sensor. It has the singular virtue of not being spoofable from a laptop.

Where the rule actually stands

Worth stating clearly, because the trade press keeps getting it wrong: there is no final BVLOS rule. I checked the Federal Register's document API on 26 August 2026. The newest action of record in docket FAA-2025-1908 is document 2026-02649, published 10 February 2026 at 91 FR 5880, and it is a proposed rule, reopening the comment period and denying a request to extend it. The original NPRM (2025-14992, 90 FR 38212) published 7 August 2025 with comments due 6 October 2025. Everything since has been comment-period housekeeping. Part 108 exists today as a proposal, not as law, and articles announcing that the FAA has finalized it are wrong.

The deadlines are worth recording because they show how long this has been late. The FAA Reauthorization Act of 2024, enacted 16 May 2024, required an NPRM within four months (by 16 September 2024) and a final rule 16 months after the NPRM published. A June 2025 executive order then required a proposed rule enabling routine BVLOS operations by 6 July 2025. The NPRM arrived 7 August 2025. The rulemaking itself started in 2021, and the advisory committee delivered its final report on 10 March 2022. Until a final rule exists, the FAA told GAO it will keep handling BVLOS through waivers and exemptions, one application at a time.

GAO made exactly one recommendation, that the FAA develop and begin implementing specific actions, including clear federal and nonfederal roles, estimated costs and technical milestones, to ensure drones can communicate with and detect and avoid other aircraft in a future information-centric national airspace system. The Department of Transportation concurred. As of GAO's June 2026 update, the recommendation is still open.

What to ask

If you are evaluating a BVLOS operation, buying detect-and-avoid capability, or reading a press release about autonomous delivery, three questions separate the claim from the capability.

First, ask which aircraft the system can see, not how far it can see. Detection range is the number vendors publish because it is flattering and easy to measure. Coverage of the non-cooperative population is the number that determines whether the system works, and it is usually absent. "Detects aircraft at 5 miles" and "detects aircraft that broadcast, at 5 miles" describe very different products.

Second, ask what happens when the broadcast is present but wrong. Every cooperative system inherits the integrity of whatever produced the data, and here that is GPS. A system that treats a received position as ground truth has no way to distinguish a correct broadcast from a spoofed one, and the FAA has now said in a GAO report that this is a scenario worth designing against.

Third, ask who is expected to yield, and whether they can. That question is not pedantry. It is the load-bearing assumption of the proposed rule, and the aircraft it assigns the duty to are precisely the ones with no electrical system, no requirement to broadcast, or an authorization to stay silent. A regulation can assign right-of-way to whoever it likes. It cannot assign perception.