Merlin Labs and the autonomous flight AI that wants to turn every airplane into a self-flying machine

Merlin Labs has flown a Beechcraft King Air 350 at FL250 with no human touching the controls — not as a drone or remotely piloted vehicle, but as a full-size turboprop governed by an onboard AI system capable of seeing, interpreting, and flying autonomously from takeoff through landing. This Boston-based company, founded in 2018 by Matt George, isn’t building a new airplane. It’s building a brain that retrofits into aircraft that are already certified and already flying.

Why Does the Retrofit Approach Matter?

Every eVTOL startup and electric air taxi venture requires a brand-new aircraft certification — a process that takes years, sometimes a decade. Merlin’s strategy sidesteps that entirely. They take an airplane the FAA already trusts, like the King Air, and add an autonomy system to it. The airframe is proven. The engines are proven. The only new element is the pilot, and that pilot is software.

This is a fundamentally different bet than what most autonomous aviation companies are making, and it gives Merlin a potential speed advantage in reaching operational deployment.

What Can the Merlin System Actually Do?

Merlin’s autonomy stack is not a fancy autopilot. A conventional autopilot follows commands — it holds a heading, holds an altitude, tracks a localizer — but it doesn’t make decisions. When something unexpected happens, it waits for the human.

Merlin’s system is designed to handle the full mission: taxi, takeoff, cruise, descent, approach, landing, and everything that goes wrong in between. It uses a combination of:

Computer vision
Lidar and radar
A cognitive AI layer that interprets situations and makes decisions the way a pilot would

The system doesn’t just follow a flight plan. It responds to weather, traffic, system anomalies, and — critically — air traffic control instructions. Merlin has demonstrated its system receiving, interpreting, and complying with ATC instructions in real airspace, with real controllers talking to a real airplane that had no human at the controls. These demonstrations occurred in the National Airspace System with FAA awareness and military authorization.

Why the King Air?

The King Air is a workhorse with thousands in service worldwide. The military operates hundreds of variants for cargo, passenger transport, and intelligence, surveillance, and reconnaissance (ISR). If you can make a King Air fly itself, you unlock autonomy for a massive installed fleet without building a single new aircraft.

That logic attracted the United States Air Force. Merlin secured Department of Defense contracts to demonstrate autonomous flight capabilities on military aircraft. The military faces the same pilot shortage as the civilian world, and many missions — cargo, resupply, ISR — don’t necessarily require a human in the seat if the technology proves trustworthy.

Merlin has flown over 100 autonomous missions, including what they describe as end-to-end flights with zero human intervention from takeoff to landing.

What Are the Hard Problems?

Certification

Even though Merlin retrofits existing aircraft, the autonomy system itself needs FAA approval. No existing certification pathway exists for a fully autonomous pilot replacement in civil aviation. The FAA has frameworks for reduced crew operations and advanced autopilots, but a system that eliminates the human pilot entirely is uncharted regulatory territory. Merlin is working with the FAA to help define that pathway, but creating a new certification category is neither fast nor simple.

Public Trust

The self-driving car industry has demonstrated this challenge clearly. Even if autonomous systems prove statistically safer than human performance, the public and regulators hold them to a different standard. A single accident involving a pilotless cargo plane — even if unrelated to the autonomy system — could set the industry back years.

Edge Cases

Aviation is full of situations that don’t appear in training data: a flock of birds at 200 feet on departure, an ambiguous ATC instruction that contradicts a NOTAM, a runway shortened by construction with markings that don’t match the database. Human pilots handle these through judgment, experience, and sometimes instinct. Teaching an AI to handle every possible combination of unusual circumstances remains an extraordinarily difficult engineering challenge. Progress is real, but “solved” and “solving” are very different words.

What’s the Go-to-Market Strategy?

Merlin’s approach is deliberately incremental:

Military cargo (current phase)
Civilian cargo operations — no passengers means a lower regulatory bar, and the economic case is strong. Cargo operators in remote or austere environments struggle to recruit and retain pilots.
Reduced crew passenger operations — not pilotless airliners, but single-pilot operations where Merlin acts as the other crew member, handling workload, monitoring systems, and capable of taking full control if the human becomes incapacitated.

That final step — the AI copilot concept — is something airlines and regulators are already warming to because it addresses the pilot shortage without asking passengers to board a plane with no pilot at all.

How Does Merlin Compare to Competitors?

Reliable Robotics takes a similar retrofit approach but uses the Cessna Caravan and keeps a remote human operator in the loop at all times. Xwing has also demonstrated autonomous Caravan flights focused on cargo.

What distinguishes Merlin is the scale of the aircraft they’re targeting and the depth of their AI. They aren’t automating simple point-to-point cargo runs. They want their system to handle the full complexity of mixed airspace alongside human-piloted aircraft. That’s a harder problem, but if solved, the addressable market is enormous — every turboprop and business jet in the world becomes a potential customer for their autonomy kit.

What This Means for Pilots

This technology isn’t replacing pilots tomorrow. But it will reshape what the job looks like over the next 15 to 20 years. The trajectory is clear: more automation, more autonomy, more AI decision support. Pilots who understand these systems, who can work alongside them and supervise them effectively, will thrive. Those who resist engaging with the technology will find themselves increasingly out of step with the industry’s direction.

Fully autonomous commercial passenger flights are likely at least 15 years away from approval in any major jurisdiction. But single-pilot operations with an AI copilot could arrive in 5 to 7 years, and Merlin’s technology is directly applicable to that use case.

Funding and Credibility

Merlin raised over $100 million in funding before going relatively quiet on the press front — a pattern that tends to correlate with companies doing real work rather than press tours. Their accumulation of actual flight hours, real operational data, and Department of Defense experience gives them a credibility advantage that flashier startups lack.

Key Takeaways

Merlin Labs retrofits existing certified aircraft with an autonomous AI system, avoiding the years-long process of certifying an entirely new airframe
Over 100 fully autonomous missions have been completed, including end-to-end flights with zero human intervention on King Air turboprops
No FAA certification pathway currently exists for fully autonomous pilot replacement in civil aviation — Merlin is helping to define one
The strategy moves from military cargo to civilian cargo to reduced-crew passenger ops, with AI copilot applications potentially 5 to 7 years out
Pilots should engage with autonomy technology now — understanding and supervising these systems will be a core competency in the next generation of aviation careers