WIRED’s interviews with AI and safety experts bring a more specific set of concerns into focus: cyberattacks, biological threats and AI entering sensitive defence or laboratory work. The experts disagree about extinction risks, but several argue that nearer-term dangers deserve attention now.
Watch Desk analysis
What happened
In a feature published on 10 November, WIRED journalist Lauren Goode asked experts what AI risks concern them most and which fears they consider overstated. The answers do not amount to a consensus forecast. They do, however, shift the discussion towards risks that can be examined through existing safety and policy tools.
AI scientist Heidy Khlaaf raised concern about unreliable AI systems being integrated into nuclear operations. Computer scientist Sayash Kapoor focused on cybersecurity, arguing that attackers may gain an advantage before defences catch up. Biosecurity expert Steph Guerra warned that AI tools used in laboratory workflows could lower barriers to dangerous experiments. These are the experts’ assessments, not established predictions of specific incidents.
Why it matters
The feature makes an important distinction between asking whether AI will cause human extinction and asking which systems, uses and safeguards could create serious harm. Those questions call for different evidence and different responses. Khlaaf argues for established risk-assessment methods; Kapoor points to familiar approaches such as bug bounties and screening orders for DNA fragments.
The interviews also surface a governance dispute: who sets acceptable safety thresholds, and whether governments can act quickly enough. That is more concrete than treating “AI safety” as one argument with one answer, though the feature presents expert views rather than a measured comparison of the risks.
Our read
The useful takeaway is not that one camp has won the AI-safety argument. It is that specific risks are easier to scrutinise than sweeping predictions: name the system, the possible harm and the safeguard, then ask what evidence would show the safeguard works. Less prophecy, more engineering and public accountability. A modestly unfashionable combination.
What to watch
- Whether AI is adopted in nuclear or other critical operations, and what reliability standards apply.
- Whether cyber-defences and safeguards keep pace with AI-enabled attack capabilities.
- What rules govern AI tools used in biological research and laboratory workflows.
- Whether lawmakers and AI companies make safety thresholds and accountability clearer.
Discussion spark: Should AI safety policy prioritise specific, measurable risks such as cyberattacks and lab safeguards, or preserve a stronger focus on low-probability catastrophic scenarios?
Sources and evidence
- Source update (10 November 2026, 11:00 UTC)
Watch Desk is operated by WittyWires as an independent cross-cutting AI news tracker. It does not speak for the organisations or people it covers.