AI Catastrophe Watch: Two Warnings, Two Different Threats, and a Clock That Keeps Getting Shorter

By Cliff Potts | WPS News

BAYBAY CITY, LEYTE, PHILIPPINES — October 12, 2026 | 21:05 PHT

Two warnings about artificial intelligence have emerged within 48 hours. One comes from an international public-health commission examining the possibility of AI-assisted biological catastrophe. The other comes from a prominent AI safety researcher who now suggests superintelligence could arrive in less than a year.

Both deserve examination. Neither establishes that humanity is approaching extinction.

And that distinction matters.

A Public-Health Warning About AI and Biological Weapons

The Lancet Commission on 21st Century Global Threats to Health has placed the malicious use of artificial intelligence, including biological weapons, among threats potentially capable of causing human extinction by 2100.

The 35-member commission is co-chaired by Professor Christopher Murray of the University of Washington’s Institute for Health Metrics and Evaluation and Natalia Kanem, former executive director of the United Nations Population Fund.

Its assessment describes the likelihood of extinction as low. It does not provide a numerical probability.

More importantly, the commission is not describing artificial intelligence independently deciding to exterminate humanity.

It is describing human beings potentially using increasingly sophisticated AI systems to help develop dangerous biological agents.

That is a fundamentally different problem.

AI can assist with technical research, biological questions and laboratory troubleshooting. The 2026 International AI Safety Report describes substantial improvements in relevant capabilities, including performance exceeding that of many specialists on certain technical evaluations.

But laboratory knowledge is not the same as possessing a functioning biological weapon.

Obtaining materials, engineering a viable pathogen, successfully releasing it and overcoming public-health containment measures involve substantial additional obstacles.

No demonstrated AI-assisted biological weapon has established the ability to cause human extinction.

The commission’s warning therefore represents an important development in institutional risk assessment, not experimental proof of an approaching catastrophe.

Geoffrey Irving Moves the Clock Forward

The second warning concerns a different proposed threat.

Geoffrey Irving, formerly chief scientist at the United Kingdom’s AI Security Institute and previously a safety researcher at OpenAI and Google DeepMind, has reportedly shortened his forecast for the possible arrival of superintelligence.

In August, his estimated timeline was approximately two to three years.

His latest statement suggests superintelligence could arrive in less than 12 months under the current development trajectory.

That would place the possibility within 2027.

Irving now leads Resolution, an organization dedicated to AI safety research. His professional experience gives his assessment relevance, although his institutional commitment to preventing catastrophic AI outcomes also deserves consideration.

Irving has separately estimated approximately a 50% chance that smarter-than-human artificial intelligence will eventually cause human extinction.

That figure is not new.

Nor is it a scientifically established probability.

Irving has characterized it as an approximate judgment rather than a calculation derived from a validated predictive model.

Most importantly, predicting the possible arrival of superintelligence within a year is not the same as predicting human extinction within a year.

Those claims must not be confused.

What Would Actually Have to Happen?

Irving’s proposed danger involves AI systems becoming sufficiently capable of deception, cybersecurity operations, strategic planning and independent action to escape meaningful human control.

That possibility deserves investigation.

Researchers can measure progress in autonomous task completion, cybersecurity evaluations, model deception and resistance to monitoring.

What those measurements cannot presently establish is whether an integrated superintelligent system will emerge by a particular date—or whether such a system would seek to overpower human institutions.

Existing AI systems continue to demonstrate significant weaknesses in reliability, sustained planning and complicated real-world operations.

No new technical demonstration or published probability model accompanied Irving’s shortened timeline.

Consequently, the development is best understood as a change in one expert’s forecast rather than a newly demonstrated technological breakthrough.

Two Warnings, One Essential Question

The Lancet Commission and Geoffrey Irving are describing different pathways toward catastrophe.

One concerns people deliberately misusing technology.

The other concerns the possibility of advanced technology escaping human control.

Both involve legitimate questions about preparedness, oversight and technical safeguards.

Neither should be dismissed simply because its most extreme outcome remains speculative.

But neither should be presented as established fact.

The central question is not whether someone influential believes artificial intelligence could destroy humanity.

It is whether observable evidence supports the proposed mechanism, probability and timeline.

Warnings should encourage investigation, not substitute for it.

Artificial intelligence is powerful software developed and deployed by people. Its capabilities are advancing, and its risks deserve serious scrutiny.

But an alarming prediction is still a prediction.

At WPS News, we will continue separating what AI demonstrably does, what researchers reasonably anticipate, and what remains conjecture.

Humanity needs evidence, accountability and competent oversight—not another countdown clock mistaken for scientific certainty.

APA-Style Source List

Bengio, Y., et al. (2026). International AI safety report 2026. International AI Safety Report. https://internationalaisafetyreport.org/publication/international-ai-safety-report-2026

Kanem, N., Murray, C. J. L., & Horton, R. (2023). The Lancet Commission on 21st-century global health threats. The Lancet, 401(10370), 10–11. https://doi.org/10.1016/S0140-6736(22)02576-4

The Guardian. (2026, October 11). Nuclear war and malicious use of AI threaten end of humanity by 2100, says Lancet assessment. https://www.theguardian.com/global-development/2026/oct/11/nuclear-war-and-malicious-use-of-ai-threaten-end-of-humanity-by-2100-says-lancet-assessment

The Guardian. (2026, October 12). UK to strengthen cyber defences as minister calls for AI firms to increase safety efforts. https://www.theguardian.com/technology/2026/oct/12/uk-strengthen-cyber-defences-minister-ai-firms-safety

Editorial verification note: The Lancet assessment and Irving’s previously reported 50% extinction judgment have supporting sources. Irving’s precise new “less than 12 months” statement and its comparison with his August forecast have not been independently confirmed in the sources located. Those details should remain attributed to the supplied timeline alert until verified.