The Growing Debate Over AI Liability
Hyderabad: Besides the chatbot on your phone, it’s no secret that AI is finding its way into countless places where decisions change freedom, security, and life expectancy, including courtrooms, airplane cockpits, hospitals, police stations, and more. These systems are evolving rapidly, but there is no clarity on who will respond when they fail.
Human beings have always relied on tools and often become overly dependent on them. Errors in such important areas do not allow for random experiments. In a courtroom in New Zealand recently, Judge Tom Gilbert discovered that apology letters submitted in an arson case were written with the help of artificial intelligence.
Regret could lead to a reduced sentence, but the judge questioned whether computer-generated remorse was a personal moral reckoning. Research by University of Kent psychologist Jim Everett, based on studies involving nearly 4,000 participants, shows that people tend to view AI-assisted writing as less original and less trustworthy. In this case, the defendant received only a partial share of remorse. Efficiency has exceeded the limits of human judgment.
In aviation, the scenario is different and bigger: Lives are at stake. As seaplane operator Capt. Augustine Joseph explains, commercial airliners rely on autopilot for most of the flight: “Once the aircraft reaches about 500 to 300 meters, the autopilot usually kicks in.” Pilots program the altitude, speed and route. They continue to update the system whenever air traffic control issues new instructions. Captain Joseph stated that advanced landing systems can direct the plane to the runway without the pilot having to touch the controls.
“Artificial intelligence is not used much in the flight part yet, but it will be used soon.” He saw a gradual shift in predictive maintenance, crew management and traffic planning before greater reliance on artificial intelligence in the cockpit, but argued that change would be slow. “Accountability and responsibility are big questions. If something happens, who’s responsible?”
This question becomes important when human instinct conflicts with system advice. “In the early stages, this needs to be a collaborative process between humans and AI.” If verification is achieved over time, pilots can be instructed to follow system guidance. Regulators will act as gatekeepers and responsibility will rest with authorities, airlines and manufacturers.
Captain CJ Chandrasekhar, director of Sky Choppers Logistics, offered a more cautious view. “I don’t think artificial intelligence will be successful in civil aviation or commercial passenger aircraft, where lives are at stake,” he said. He has accepted a role in ground handling and cargo logistics, perhaps drones. “But not in passenger operations or aircraft operations. Absolutely not.”
His concern was obvious. “How safe is artificial intelligence? Who is responsible if something happens?” The same liability issue comes back.
Additionally, artificial intelligence systems learn from historical data. In India, researchers have warned that models trained on social stereotypes can internalize caste hierarchies where surnames and professions become proxies. Hyderabad-based software engineer Kiran R. pointed out that personalization is an everyday example.
“Just like your Instagram feed differs from someone else’s, AI’s information and decisions vary depending on who you are,” he said. “A man and a woman may be shown different content. Someone from one class may see different news than the other.”
The best example of this when it comes to military use is AI company Anthropic’s recent public statement that the US government rejects practices such as mass domestic surveillance and fully autonomous weapons, even if they are legally permitted. They argued that mass surveillance threatens democratic freedoms and that current AI systems are not reliable enough to power weapons without human control. But the Pentagon has signaled it will only contract with companies that agree to “any lawful use.”
Health and policing follow similar paths. When a diagnosis is missed or a particular class neighborhood is overpoliced, the path to accountability passes through data, software, corporate policy, and human oversight. Courts then evaluate the results long after the design choices that produced them.
If an AI system one day recommends a punishment, flags a suspect, guides a plane or filters a medical test and something goes wrong, who will respond? Is it the programmer who writes the code, the company that sells it, the official who approves it, the operator who implements it?
