Two flights, a few months apart, show exactly where AI and automation stand on the flight deck today.
On 20 December 2025, a Beechcraft King Air climbing through FL230 after departing Aspen suffered a rapid loss of pressurisation.
The two pilots pulled on their oxygen masks.
Then they let the aircraft land itself at Rocky Mountain Metropolitan Airport near Denver, in what Garmin confirmed was the first real world activation of its Emergency Autoland system, six years after its 2019 debut.
In May 2026, CNN reported from the right seat of a Merlin Labs Cessna Caravan as it departed Quonset State Airport in Rhode Island.
The test pilot kept his hands off the controls. An AI system flew the takeoff, listened to a simulated air traffic controller and replied over the radio.
Flights like these keep one question alive in crew rooms, flight schools and departure lounges. Will AI remove the need for pilots?
Our answer, based on the regulatory record, the technology and the economics, is no. Not on passenger airliners, and not any time soon.
AI is already reshaping the job, however, and the first pilotless commercial operations are taking shape where there are no passengers to reassure.
Automation Is Not the Same as AI
Airliners have relied on automation for decades.
Autopilot, autothrust, autoland and fly by wire logic already handle much of a routine sector.
What matters is how those systems make decisions.
Conventional automation follows rules that engineers write, test and certify.
AI learns its behaviour from data.
Stanford researcher Mykel Kochenderfer told CNN that today’s cockpit automation operates within tightly defined limits, while newer AI is designed to cope with a much wider variety of surprises.
Garmin’s Autoland is best understood as advanced conventional automation, and its first real rescue carried a lesson.
The system triggered automatically as cabin altitude rose past safe levels, and the crew then made a deliberately cautious decision to let it finish the job.
Its automated radio calls, however, told controllers the pilots were incapacitated, which fed incorrect early reports.
The humans on board understood the situation. The software could only follow its script.
Airbus has gone further with genuine machine learning.
Its ATTOL project used onboard image recognition to achieve autonomous taxi, takeoff and landing on an A350-1000 in 2020, after more than 500 test flights.
Airbus said the aim was to let pilots spend less effort on handling the aircraft and more on strategic decision making.
Its later DragonFly demonstrator simulated a sudden pilot incapacitation over Limoges, and the A350 diverted itself safely to Toulouse.
Airbus told FlightGlobal the work was designed to support a two pilot flight deck, not a single pilot one.
Europe Put the Brakes on Single Pilot Flying
The real fight over the flight deck in recent years has not been about empty cockpits. It has been about reduced crew operations.
Extended Minimum Crew Operations, or eMCO, would leave one pilot alone at the controls during cruise while the other rests.
Single Pilot Operations, or SiPO, would remove the second pilot for the whole flight. Airbus and Dassault first approached EASA about eMCO in 2021.
For a while, the momentum looked unstoppable.
In 2022, Bloomberg reported that more than 40 countries, including the UK, had asked ICAO to help develop a safe path to single pilot flights.
EASA suggested such services could begin in 2027.
Then the evidence arrived.
In June 2025, EASA published the final report of research carried out by a consortium led by the Royal Netherlands Aerospace Centre (NLR).
It concluded that a level of safety equivalent to normal two pilot operations “can currently not be demonstrated”.
Before it reconsiders, EASA wants flight decks with proven systems for managing workload, monitoring pilot health, spotting security threats and providing automated safety backups.
Its rulemaking will now concentrate on developing and proving that technology inside today’s two pilot flight decks.
The research also exposed problems that software alone cannot fix.
It examined physiological needs such as lavatory breaks, which could leave the controls unattended for extended periods, and IFALPA says the work foresaw no mitigation for decades.
Pilot groups carried those findings to ICAO’s 42nd Assembly in Montreal.
A paper from IFALPA, the ITF and IFATCA argued that eMCO rests on productivity and labour cost arguments rather than any existing safety problem.
It added that successful AI development in aviation remains uncertain even for the most basic applications.
Manufacturers read the logic differently.
Christian Scherer, speaking as chief executive of Airbus’s commercial aircraft business, argued that the technology is feasible and that once a flight deck is down to one pilot, removing the last one becomes the logical next step, AeroTime reported.
That is precisely why unions fight eMCO so hard.
Washington Writes Two Pilots Into Law
The United States has moved from caution to legislation.
In July 2025, US senators urged the FAA to oppose single pilot operations at ICAO.
In February 2026, Congress passed a spending package that ended a government shutdown.
According to ALPA, it directs the FAA to ensure a minimum of “two qualified, fully rested pilots on commercial airline flight decks at all times.”
The provision arrived through an appropriations bill rather than a standalone aviation act. The signal to manufacturers is unambiguous all the same.
Europe’s regulator has paused single pilot rulemaking, and American law now anchors the flight deck on two pilots.
The AI Roadmaps Point to Teamwork, Not Replacement
Regulators are not ignoring AI. They are planning for it in stages.
EASA’s AI Roadmap 2.0 sets out three levels. Level 1 assists the human.
Level 2 covers teaming between humans and AI, which EASA expects around 2035 and which could include systems that support the pilot flying during reduced crew operations.
The agency stresses that the date is a prediction, not a target.
Level 3 covers advanced automation, first with human supervision and ultimately without it.
EASA places that shift progressively between 2035 and 2050, and expects commercial air transport to take a stepped approach.
Its first regulatory proposal on AI, opened for consultation in late 2025, focuses on Level 1 assistance and Level 2 teaming.
Nobody is writing certification rules for an AI captain.
The FAA takes a similar line.
One guiding principle in its Roadmap for AI Safety Assurance is to avoid personification, treating AI as a computer system rather than a human colleague.
Closer to home, the UK CAA’s strategy for regulating AI says any AI involved in controlling an aircraft must be proven to decide correctly every single time.
It also flags a second challenge: making sure people and AI work well together.
What Pilots Do That Algorithms Still Cannot
On 4 November 2010, Qantas flight QF32 was climbing through 7,000 feet out of Singapore when the A380’s number two engine suffered an uncontained failure. Debris punctured the wing and damaged around 650 wires.
The autothrust failed to hold the target speed, so the captain pulled the thrust back by hand. The ECAM filled with warnings.
Working through the first wave of procedures alone took about 50 minutes.
Then came the improvisation.
Heading into the cabin to inspect the damage, the second officer found help from an unexpected source: a passenger showed him the tail camera feed on the inflight entertainment system.
Fluid was pouring from the left wing, and a check from the lower deck confirmed it was fuel.
The crew landed the badly damaged A380 back at Changi, and nobody on board was hurt.
That is the heart of the case for human pilots.
Aviation’s worst days rarely match a checklist, and they are exactly the events that are scarce in any training dataset.
A landmark 2013 FAA and industry report found that pilots routinely catch and correct problems, noting that “the aviation system is designed to rely on that mitigation.”
The same report also found weaknesses in manual flying as automation took over.
An update to that work, cited in a paper to ICAO’s 40th Assembly, found manual flight errors were causal or contributory in 92 per cent of flight path management accidents worldwide between 2009 and 2016.
AI developers seize on figures like that.
Merlin chief executive Matthew George told CNN that human error still causes around 80 per cent of aviation accidents.
Both sides have a point, which is why regulators favour pairing pilots with better tools rather than choosing one over the other.
Passengers Are Not Ready to Fly Without a Pilot
Even with mature technology, airlines still need people to buy tickets.
In 2017, UBS estimated that pilotless aircraft could save the industry $35 billion a year.
Its survey of 8,000 people across the US, UK, France, Germany and Australia found 54 per cent were unlikely to take a pilotless flight. Only 17 per cent said they would.
A 2018 Ipsos survey for ALPA found that 81 per cent of American adults would feel uncomfortable on a completely pilotless airliner.
Just 15 per cent said they would be comfortable.
Those numbers are close to a decade old, and younger respondents were noticeably more open.
But no airline wants to test public trust with its own brand.
Where Pilotless Flight Will Arrive First
UBS predicted that cargo would move first.
The investment trail agrees.
In May 2026, Nasdaq listed Merlin launched Merlin Pilot for Commercial Cargo and confirmed it is collaborating with freighter lessor World Star Aviation.
The same autonomy core is progressing through military airworthiness with US Special Operations Command on the C-130J.
Merlin insists it is building a partner, not a replacement.
“We’re not flipping a switch to uncrewed airplanes,” George told CNN. Yet the company has also described an incremental path toward passenger flights without pilots, with “no hard ceiling” on aircraft size.
Reliable Robotics raised $160 million in April 2026 and is pushing for FAA certification in 2028.
Back in November 2023, it flew a Caravan with nobody on board, supervised from about 50 miles away.
Boeing subsidiary Wisk flew its Generation 6 autonomous air taxi for the first time in December 2025 and added a second test aircraft in May 2026, with commercial service targeted for 2030.
There is no pilot on board. Instead, a supervisor on the ground oversees several aircraft at once.
Military aviation is further ahead again.
In September 2023, an AI agent flew the X-62A VISTA in dogfights against a human flown F-16.
Safety pilots on board never needed to disengage it, and then Air Force Secretary Frank Kendall flew in the jet in May 2024.
Look closely and a pattern emerges.
Pilots are not vanishing from these operations. They are moving to the ground, into supervisory roles and, eventually, towards one person watching several aircraft.
The Pilot Shortage Argument Cuts Both Ways
Supporters of reduced crew flying often point to the pilot pipeline.
Boeing’s 2026 Pilot and Technician Outlook forecasts demand for 674,000 new commercial pilots over the next 20 years.
Boeing says around two thirds of all the aviation hiring it forecasts will simply replace people retiring or leaving.
Fewer pilots per flight would ease that pressure.
IFALPA counters that the flight deck is where experience passes from one generation to the next.
Splitting crews for long stretches of a flight, it argues, weakens that mentoring just as thousands of seasoned pilots retire.
For now, AI’s most practical contribution sits inside the training pipeline itself.
Boeing has highlighted AI alongside virtual and mixed reality as tools to strengthen how the next generation learns to fly.
Our Verdict on AI and the Flight Deck
AI will not remove the need for pilots on passenger airliners any time soon.
Even EASA’s own roadmap does not place advanced AI automation in commercial air transport before 2035, and it expects that shift to unfold gradually towards 2050.
The likely sequence is already visible.
AI assistance comes first. Teaming between pilots and AI follows.
Uncrewed operations begin with freight and air taxis, supervised from the ground.
Meanwhile, single pilot airliners remain paused in Europe and blocked in the United States.
What will change is the job itself. Pilots will spend more time supervising smarter systems and less time flying by hand, which makes manual handling training more important, not less.
For anyone starting flight training today, Boeing’s forecast of 674,000 new pilots is a far more useful number than any AI timeline.
AI is learning to fly. Earning the trust to fly alone is a much longer journey.
Continue to follow The Aviation Hub for more analysis and insight!
