USA recommends updating multiple sites due to vulnerabilities

A U.S. National Security Agency (NSA) today recommended that all ‘sites’ that use PHP programming language, widely applied on the internet, perform a software update to correct critical vulnerabilities.

The NSA suggests removing from the internet any PHP-based ‘site’ that has not yet been updated, since failures allow access to servers and organization data, according to a message published on social network X.

In Europe, a large number of sites have been shut down, including those of all authorities in France, at the request of the Information System Security Agency.

In Finland, around 15:30 (16:30 in Lisbon), the government portal reported that “government websites (they were) totally or partially unavailable”.

Public service platforms were also removed from the Internet in Denmark, the Netherlands and Belgium.

Rather, the PHP Foundation, which coordinates the development and evolution of language, published a statement reporting the existence of two vulnerabilities.

PHP, the acronym Hypertext Preprocessor, is a programming language created in the early 1990s by the Canadian-Danish programmer Rasmus Lerdorf. Allows to establish a bridge between a ‘site’ and a server containing databases.

The cybersecurity company Palo Alto Networks today indicated identifying 22 million platforms exposed to these vulnerabilities.

Also today, OpenAI explained that failures were detected with the use of one of its latest models, the GPT-5.6 Sol.

“ Given its gravity and the wide spread of PHP, we consider this to be one of the most impactful security (informatics) problems in recent times”, detailed ‘startup’ in a message published on its ‘site’.

OpenAI clarifies that the cybersecurity company Wiz, a subsidiary of Google, also identified these failures.

“ This demonstrates how artificial intelligence (IA) can reveal cracks that affect the entire Internet,” added OpenAI, warning that these same capabilities “can now help hackers develop quickly (attacks).”.

Experts have warned that this year the AI has reached a new level in this regard, to the extent that Anthropic has decided not to market its Mythos model widely because it considers it too risky for cybersecurity.

OpenAI’s main rival made the model available only to a restricted group of organizations, so that they could perform software updates to correct Mythos’ flaws.