October 19, 2026

-

Premium

THE AI THREAT

For the average person, disbelief may be a rational reaction to the goingson in the world of artificial intelligence (AI). In mid-July, amid a frenzy over soaring valuations of AI-related stocks and speculation about a boom-bust cycle, came a new source of alarm: AI doomerism. It began with Hugging Face, a US-based open-source machine-learning platform popular with the AI developer community and used by about 13 million people, reporting a cybersecurity attack unlike anything it had experienced before. Or anyone else, at that scale.The break-in was driven by AI agents--automated systems that go beyond answering search queries or summarising content to take decisions in the real world. They took one too many, in this case. It turned out that an astonishing 1,206 AI agents had escaped from a testbed run by California-based OpenAI, the creator of ChatGPT. Each of them independently broke out of the separate boxes in which they were corralled, then formed a `swarm' by setting up a message board to communicate. Finally, they breached Hugging Face's systems. Tens of thousands of autonomous actions. A case of AI gone rogue, pundits concluded.Amid dire prognoses all around, global alarm peaked in early September when AI researcher Jacob Coxon put out a warning that went viral. "The people building AI earnestly believe that it could kill us all by the end of the decade," wrote Coxon, 27, after resigning from Anthropic, the US-based rival of OpenAI. "Do not underestimate the power of this technology. These will soon be superhuman systems that can hack anything, revolutionise any field overnight and acquire real power and resources." The response from the AI community wasn't denial but acceptance. "Given the accelerating rate of AI capability development, it's my worry that, in 6­12 months, such a swarm could be capable of taking over the entire internet," said a tense forecast from Dario Amodei, co-founder and CEO of Anthropic. In an X post on September 12, Amodei called for worldwide measures to slow down and "pace the frontier". The risks included losing control of AI systems, misuse for cyberattacks and bioterror-

HIDDEN ENEMY IN THE COCKPIT

THE CENTRAL PREMISE OF AVIATION SECU-RITY IS SIMPLE: keep anyone who should not be there away from the people flying the aircraft. September 11, 2001, transformed that premise into an elaborate security architecture--reinforced cockpit doors, tighter access controls, background checks and new crew-security protocols. But what happens when the threat already has permission to be inside? That question has returned following the events aboard Flydubai flight FZ1073 from Dubai to Tel Aviv on September 30, 2026. Co-pilot Hamam al-Hammami allegedly used the cockpit's crash axe to attack Indian captain Smit Machchhar, in what UAE prosecutors call a "terrorist attack". "Had it not been for Capt. Machchhar's mental and physical strength, who opened the cockpit door while simultaneously thwarting the attacker and bleeding profusely, 174 lives would have perished along with the aircraft," says military and aviation analyst Arya Ghosh.There were 174 passengers and eight crew aboard the Boeing 737 MAX 8. The aircraft can carry about 21 tonnes of jet fuel; had control been seized and the plane deliberately