AI Cyber Risks, Shorter AGI Timelines and the Energy Case for Electric Cars

Three shifts are converging: AI is changing the pace of software attacks, Demis Hassabis is putting a shorter timeline on artificial general intelligence (AGI), and electric cars are giving oil-importing countries another way to reduce exposure to supply disruptions. Each raises a practical question: can systems adapt as quickly as the technology changes?

Software security faces a faster threat cycle

AI is making it easier to turn small software flaws into serious vulnerabilities. That changes the economics of defense: flaws that once required time and specialist skill to find or exploit may now attract attention much faster. The internet’s openness, which has helped it grow and connect people, can also make weaknesses easier to reach.

The near-term response may involve restricting some of that openness to improve security. That could mean harder choices about access and control, with costs for the easy exchange of information and services that users expect. The point isn’t that openness must disappear; it’s that the balance between access and protection may become harder to maintain as attack methods speed up.

The longer view needn’t be fatalistic. The argument is that a period of acute risk can prompt new technical foundations, much as the pandemic spurred work on mRNA vaccines. Software and internet infrastructure could also be redesigned to earn greater trust. That outcome isn’t automatic: it depends on whether teams can make security a property of the systems they build, rather than a patch applied after flaws surface.

There’s a useful historical pattern here. Changes in how people live and work can intensify existing forms of pollution or exposure. Animal domestication brought people into closer contact with livestock and zoonotic disease. Urban growth helped diseases spread, and sanitation became part of the response. These examples don’t predict exactly what will happen online, but they suggest that new capabilities can create new burdens—and that societies can respond by changing infrastructure.

The internet’s current security pressure may be another such moment. Treating it as inevitable catastrophe could obscure practical work; dismissing it could leave systems exposed. The central test is whether security improvements can keep pace without erasing the benefits that made the network useful. Read the full essay on the internet’s security future.

Hassabis compresses his AGI estimate

Demis Hassabis has moved his estimate for AGI closer over the past year. At Davos in January, he put the horizon at five to ten years. Later, he cited three to five years. He now tells the Wall Street Journal that it could be two to three years away.

The change matters because Hassabis is a prominent AI researcher with a view of progress inside his own lab. His estimate is still a forecast, not a measured result or a settled definition. People disagree about what AGI should mean, and the label may not be useful for judging systems that acquire capabilities unevenly. Even so, the shorter timeline signals that Hassabis expects a significant threshold of ability to arrive sooner than he previously anticipated.

That forecast also points to a second constraint: discovering something with AI and putting it to work are different tasks. New capabilities must be tested, understood, and applied in settings where mistakes carry real costs. The faster systems improve, the more pressure there is to build ways to evaluate them and determine where they are dependable. A capability milestone alone won’t answer how quickly research results can be validated or adopted.

For developers and organizations, the practical question is less about the label than the specific work a system can do, and how reliably it does it. Watch for clearer demonstrations, repeatable evaluations, and evidence that new capabilities transfer beyond the setting in which they were developed. Those details will help distinguish a nearer technical threshold from a nearer path to useful deployment.

Electric cars and exposure to oil disruption

Electric vehicles can give countries that import oil a way to reduce their exposure to disruptions in oil supply. As cars shift from liquid fuel to electricity, transport demand becomes less directly tied to imported oil. That creates an energy-security benefit alongside the more familiar case for electrification.

The benefit depends on the wider energy system. Charging still requires electricity, and the effect on oil demand depends on how many vehicles switch and how much they are driven. Electrification therefore changes the source of exposure; it doesn’t make transport independent of energy supply. The useful question for each country is how much oil use electric cars can displace and how resilient its electricity system is.

This framing makes the transition relevant even where climate goals aren’t the only concern. Reducing reliance on one fuel can give importing countries more options when supply is disrupted. The scale of that advantage will depend on adoption and the available electricity, so the security dividend should be judged through actual changes in transport fuel use.

Ideas, readers and the next questions

Bill Gross, founder of Idealab, has launched more than 175 companies and pioneered paid search through GoTo/Overture. His latest venture, Idealab.ai, draws on that experience to identify business opportunities and make the ideas freely available. He’ll discuss the platform and take questions at an online salon for Exponential View members on November 5, at 8am PT and 11am ET. Find event details.

The issue also flags a Russian satellite’s close encounter, attempts to turn mercury into gold, and plastics designed to self-destruct. Readers are being invited to complete a short survey about possible chat rooms for meeting other subscribers and exchanging ideas.

The threads across these topics are speed and adaptation. AI is accelerating the discovery and exploitation of software flaws; Hassabis’s forecast suggests expectations for advanced AI are moving closer; and electric cars can alter how much transport depends on imported oil. In each case, the payoff depends on the systems built around the technology: security practices, evaluation methods, and energy infrastructure.