Fridays, 2.30 pm (Lisbon) · Artificial intelligence on NOW →Fridays, 2.30 pm · AI on NOW →
Miguel Sousa SoaresMSS
← Media
NOW · Weekly segment · 30 October 2026

Attacking Has Become Cheaper Than Defending

A drone costs thousands of euros. The missile that brings it down costs millions. In the digital world, artificial intelligence is doing the same sums: attacking has become cheap; defending is still expensive.

Miguel Sousa Soares joining remotely for NOW’s weekly artificial intelligence segment, 30 October 2026

Original broadcast in Portuguese · NOW · Weekly segment · 30 October 2026 · 13 minutes · Watch on YouTube

“A promotional deepfake is dangerous.”

The new economics of cyberattacks

Two critical flaws in PHP, widely used on websites, prompted organisations to suspend services as a precaution. The Netherlands’ NCSC confirmed that it temporarily took its own website offline and that other organisations, including public agencies, did the same.

  • Signs of one problem were known in 2024, but concrete exploitation vectors had not been found. In 2026, researchers demonstrated remote code execution. The PHP Foundation describes how AI-assisted research helped discover and validate flaws; this does not establish that AI caused the observed attempts.
  • AWS identified one exploitation attempt dated 17 October, four days before disclosure. On 21 October it observed several proof-of-concept exploits and further attempts, without knowing whether they would succeed against real-world systems.
  • Research, demonstrating a vulnerability and compromising a system are different events. Taking a website offline as a precaution is not evidence that it was successfully attacked.
  • The asymmetry remains: attacking means looking for an open door. Defending means discovering, checking and fixing it without opening another.

Topic 2 · Who informs the voter?

With the US midterm elections only days away, more and more people are asking chatbots whom to vote for or what each candidate stands for.

In a Washington Post test of six chatbots, almost half the election guides were incomplete or wrong. One put a non-existent Senate election on the ballot.

We can have secure voting systems and, at the same time, unreliable information influencing voters’ decisions.

The greatest risk is the promotional deepfake: it aligns a candidate’s posture, micro-expressions and message, and can be more persuasive than the candidate themselves when they hesitate.

Weekly artificial intelligence segment

Fridays, 2.30 pm (Lisbon), on NOW.

All broadcasts

Miguel Sousa Soares is an economist and analyst of geoeconomics and artificial intelligence, with more than twenty years’ experience in strategy and corporate finance. He writes for Visão, Jornal de Notícias, Expresso and Correio da Manhã. He comments on artificial intelligence on NOW on Fridays.