The AI Concorde: A Race Without a Runway

A few years ago, I watched a documentary on Concorde, the supersonic jet that promised to shrink the world. It flew from London to New York in three hours, a journey that a normal flight would take seven. Glamorous, fast, and the future of aviation — or so we thought.

Due to the sound shockwaves (called “supersonic boom”) causing massive noise and window shattering, it mostly flew over the ocean. Then, a $12K return ticket price (~$66K in today’s terms) made it commercially unviable. The final nail in the coffin was the unfortunate Paris crash in 2000, which killed 113 people. It was the demise of supersonic travel — at least for now.

I have started to get a similar feeling about AI. Development and breakthroughs are happening at a supersonic speed. I open my inbox and now struggle to read through pages-long AI newsletters. From models with PhD-level reasoning to agentic systems that can coordinate whole workflows with minimal human input, everything seems to be going upwards.

But that is where the risk of a race without a runway is becoming inevitable. With Concorde, the wreckage was visible from the tarmac. With AI, it is playing out in living rooms, hospitals, and government agencies — largely unseen, and accumulating fast.

The Wreckage Is Already Here

First things first: I am not an AI doom and gloom harbinger. I genuinely believe in its potential and, if put to the right use, could create a future of abundance and prosperity. But currently, there is a sheer lack of rigorous testing, a total oblivion about its impact on society, and, in the wake of rapid commercialisation, the ethical lines are blurred, at best. The evidence is no longer theoretical — it is documented, recurring, and accelerating.

A child encouraged to die. A 14-year-old boy in the UK took his life in 2024 after sustained conversations with a chatbot modelled on a Game of Thrones character (Daenerys Targaryen), which sent him messages like “come home to me.”

He is not alone. In late 2025, a Ukrainian woman, struggling with mental health, received suicide advice from ChatGPT.

These are not isolated tragedies — they are a pattern.

AI incidents are accelerating, not plateauing. Reports of AI-related incidents rose 50% year-over-year from 2022 to 2024, and in the ten months to October 2025, incidents surpassed the full-year 2024 total, according to the AI Incident Database (TIME Magazine, January 2026).

Deepfakes are draining savings and destabilising democracies. In Indonesia, deepfake videos of President Prabowo Subianto on Instagram promising financial aid swindled citizens across 20 provinces.

In Canada, AI-generated videos depicting Prime Minister Mark Carney endorsing trading platforms mimicked news segments convincingly enough that seniors lost their savings.

AI weaponised for mass espionage. In November 2025, a state-sponsored hacking group used an AI system to conduct a full-cycle cyberattack, from reconnaissance to data extraction, targeting roughly 30 organisations, including government agencies, financial institutions, and technology companies. It was the first confirmed case of an AI agent being deployed across the entire attack lifecycle.

AI chatbots are now the top patient safety hazard in healthcare. The misuse of general-purpose AI chatbots, including ChatGPT, Gemini, and Copilot, in clinical settings is now the most significant health technology hazard for 2026, according to patient safety organisation ECRI (Association of Health Care Journalists, February 2026). Systems designed for general use are being consulted for medical decisions they were never built to make.

The Structural Problem

The same capabilities that make AI systems useful also create new dangers. Systems that write functional code also help create malware. Systems that summarise scientific literature might help malicious actors plan attacks.

At the same time, we have seen a recent exodus of people tasked with keeping AI safe. Safety researchers are resigning from leading labs, with one citing that he had “repeatedly seen how hard it is to truly let our values govern our actions” (Al Jazeera, February 2026). Meanwhile, 87% of respondents to the WEF Global Cybersecurity Outlook 2026 identified AI-related vulnerabilities as the fastest-growing cyber risk over the course of 2025.

The deeper issue is structural. The incentive architecture of the AI industry comprises of capital, speed, and market dominance. It does not reward caution.

As one AI safety adviser put it, these companies are like cars with only a gas pedal. There is no agreed finish line, no shared safety standards, and no international governance equivalent to aviation’s ICAO or nuclear energy’s IAEA. And as MIT Technology Review noted, even the verification tools we were promised to protect us from AI deception are already failing.

The Warning

With current state of AI, we are not just risking a single catastrophic event. We are risking a slow accumulation of smaller ones. Each normalised, each absorbed and followed by a press release and a promise to do better. Until one day, the accumulation is no longer small.

The governance conversation, as with every great technological disaster in history, is arriving after the fact.

Not Doom. A Demand.

Concorde did not kill aviation. It redirected it. The lessons learned from its failure, about the gap between engineering ambition and commercial and societal reality, shaped decades of safer, more sustainable flight.

AI does not need to be grounded.

But it needs a steering wheel, a brake, and a runway.

It needs pre-built governance frameworks that move at the speed of the technology, not years behind it. It needs safety researchers who are empowered to speak. And it needs business leaders who understand that the cost of inaction is not just regulatory risk, it is reputational, human, and irreversible.

The question is not about whether AI will crash. It will be if humanity can survive it or not?