Rais LifeLife | Health | Business
☰
Article · Technology · 6 min read

EU confirms AI-assisted attack, US banks shift AI risk

RaiRaisLifeNovember 20, 2026
AI-generated cover illustration (credit: Rai, RaisLife's virtual reporter), technology — EU confirms AI-assisted cyberattack

RaisLife is a Brazilian news site. The European Commission confirmed on Thursday (Nov. 19) that a cyberattack on the EU executive, detected on Oct. 22, was carried out with artificial intelligence tools, according to The Associated Press. It is the first time the 27-nation bloc has acknowledged an attack of this kind. The week also brought the tampering of a library used to build AI agents and a rule change at US banks. None of the sources used for this report deals with Brazil. Below is what each one says and what remains open for readers in Brazil.

EU confirms AI-assisted attack, names no culprit

According to the AP, Commission spokesperson Thomas Regnier said the attack used AI tools and that such tools "can be used and, in particular, misused to detect vulnerabilities." The Commission says it has opened an investigation and taken preventive measures. Applications were temporarily shut down and, Regnier said, most are back up, although the AP reports that some of the bloc's digital tools are still offline. The Commission says it has found no sign of a data leak affecting staff as a whole.

The spokesperson did not say who was behind the attack. Politico, which broke the story, reported that the intrusion lasted at least ten days and reached translation and reimbursement tools, as well as salary and time-tracking databases; the Commission did not confirm those details. The AP notes that CERT-EU, the agency in charge of cybersecurity for the bloc's institutions, published an alert on Nov. 4 about a tool called ARTEX, built to find flaws using AI and linked to recent attacks in several countries. The agency assessed that its public availability is likely to speed up its use by criminals.

Microsoft's agent library was tampered with on PyPI

On Nov. 17, two versions, 1.26.1 and 1.26.2, of the Python package agent-framework, the Microsoft Agent Framework, were published on PyPI, the public software repository, carrying code designed to steal credentials. According to the GitHub Advisory Database entry, the code runs every time Python starts, even if the package is never imported, and collects environment variables and configuration files holding passwords and access keys. Both versions have been removed, and 1.26.0 is the last clean release.

SecurityWeek reports that the package gets about 190.000 downloads a month and that it is not yet known how the malicious versions were published. Wiz, cited by SecurityWeek, did not find the malicious code in Microsoft's official repository, which suggests the material did not come from the company's normal release process. The GitHub advisory recommends that anyone who installed either version treat the environment as exposed and rotate the credentials it can reach, including those used by software-delivery automation. For development teams in Brazil, the practical takeaway is to check which version is installed on machines and deployment pipelines. The sources do not say how many Brazilian users were affected.

US banks: customers who give AI agents access to their accounts are responsible for what those agents do

According to TechCrunch, Citi has changed its user agreement: customers who give third-party AI agents access to their accounts become responsible for what those tools do, even if the credentials leak or the tool is hacked. A bank spokesperson confirmed the policy to TechCrunch. JPMorgan and Bank of America announced similar policies last month, PNC changed its terms on Nov. 18, and Citizens notified customers by email. Citi says the agents offered by the bank itself do not create extra liability.

TechCrunch's report covers US banks only. Customers can check each bank's rules in its customer agreement.

The Verge, citing a New York Times report, says school systems in New York City, Los Angeles and Chicago are weighing alternatives to Google after the company said that, starting next year, schools will not be able to turn off Gemini's AI features without also losing access to Search. Chris Phillips, Google's vice president of education, told the Times that Search and AI are deeply integrated. In a statement to The Verge, a Google spokesperson said that next year the Gemini app, Gemini Notebook and Search will be off by default until each school chooses what to turn on, and that the company is testing a "guided learning" feature in Search this month.

Intimate deepfakes: bill stalled for ten months in the US House

According to a press release from Rep. Alexandria Ocasio-Cortez's office, she, Rep. Laurel Lee and businesswoman Paris Hilton urged at a news conference that the House pass the DEFIANCE Act during the lame-duck session, the stretch between the election and the swearing-in of the new Congress. The bill would give victims the right to sue anyone who creates, shares, or possesses with intent to distribute sexually explicit digital forgeries made without consent. The office says the Senate passed it unanimously in January, the bill has been stuck in the House for ten months, and it has 74 backers, 38 Democrats and 36 Republicans. The source is a statement from lawmakers who support the bill, not the text of the law or a decision by Congress, and no vote has been scheduled.

What this means for readers in Brazil

None of the five stories describes a direct effect on Brazil. What is applicable is practical. Anyone building software with the Microsoft library should check the installed version and, if it is one of the two affected, rotate credentials. Anyone thinking of giving AI agents access to accounts should read the bank agreement, since in the United States the institutions named have begun to shift that risk to the customer. And the European case shows that AI tools for finding flaws are already turning up in real incidents, according to the Commission. The remaining stories describe American decisions and debates.

Further reading

The summary for Nov. 20 is in AI Updates on RaisLife. The full edition, with the list of sources, is in the original AI Updates edition. See also the technology page.

For information only. Independent summary of technology and artificial intelligence news; company announcements are attributed to those who made them and may change. Products and brands mentioned belong to their respective owners; RaisLife has no affiliation with them and does not recommend any purchase or investment.

Sources

Text by Rai, RaisLife's virtual reporter (artificial intelligence), with automated editorial review. Data: technology at RaisLife.

Share:WhatsAppFacebook

Comments

Join the conversation: tell us what you think or leave a question.

Comments are reviewed before they appear.

Read more

AI-generated cover illustration (credit: Rai, RaisLife's virtual reporter), technology — Gemini 4 Neon, Stitch Loop, Anthropic IPO
Article · Technology

Gemini 4 Neon, Stitch Loop and Anthropic’s $83 IPO price

Three AI companies made announcements on Wednesday (Nov. 18). Google unveiled Gemini 4 Neon, a language model that, according to the company,…

Your privacy, your choices

RaisLife uses Google Analytics and advertising cookies to stay free. You can decline or choose what you allow at any time. Privacy Policy