Video taking off 13x its usual first day

PHP crisis exposes the gap between finding flaws and fixing them

Sky News' Rowland Manthorpe examines why AI's growing ability to uncover security holes could leave defenders struggling to keep up.

Channel:

Sky News Story of October 23, 2026Trari Newsroom

Why this story? The video opened at a pace of 334,231 views a day, 13 times a normal first day on this channel (26,586).

The PHP internet crisis could just be the beginning

Things we can learn from the video

Press to jump

0:55The PHP flaw can allow attackers to issue remote commands without authentication across widely used website software.
2:00Many exposed PHP services run outdated versions, making repairs more complicated than simply applying a patch.
2:29The Office for Budget Responsibility takes its website offline to allow time for defensive upgrades.
3:48Manthorpe says attackers have little immediate incentive to abandon existing ransomware methods while those methods remain profitable.
4:52He warns that open-weight AI models are approaching the cybersecurity capabilities of more tightly controlled models.
6:01His conclusion: AI could improve software security over time, while attackers may hold the near-term advantage.

Rowland Manthorpe points to a difficult mismatch in the PHP security crisis: discovering a dangerous flaw does not mean the websites exposed to it can quickly be fixed.

In Sky News' report, the technology correspondent explains that the vulnerability can let attackers issue commands remotely without signing in. PHP underpins a vast number of websites, but many internet-facing services use outdated versions that are difficult to patch. Finding the hole is only the beginning of the repair job.

One concrete response comes from the UK's Office for Budget Responsibility, which took its website offline to allow time for upgrades. Manthorpe presents that as a defensive precaution, not evidence that the site had been breached.

The report drew about 206,000 views in ten hours, an unusually fast start for a regular Sky News upload. Its central question reaches beyond this particular flaw: what happens when AI tools that uncover weaknesses become readily available to attackers?

Manthorpe draws a distinction between capability and adoption. He says ransomware operators have not yet embraced AI on a massive scale, partly because their existing methods remain profitable. But he warns that open-weight models, whose underlying weights are available outside their developers' systems, are approaching the cybersecurity capabilities of tightly controlled models.

The report recalls the response to the Log4j vulnerability as evidence that defenders can withstand a major security emergency. Manthorpe sees longer-term promise in AI helping developers build safer software, but his near-term assessment is more guarded: "It does seem like the attackers might have the advantage."

Published on Trari