Brief IA
›

Microsoft fixes 1,413 vulnerabilities, Nadella warns about AI

Business & Startups·Tom Levy·

Microsoft fixes 1,413 vulnerabilities, Nadella warns about AI

Microsoft fixes 1,413 vulnerabilities, Nadella warns about AI
Key Takeaways
1Microsoft fixed 1,413 vulnerabilities in October, surpassing its previous annual record.
2According to the FBI, some intrusions using ARTEX achieve administrator rights in less than three hours.
3Calif discovered Java4Shell with OpenAI's Cyber GPT-6; Oracle released a patch on November 17.
Why it matters : according to Satya Nadella, already exploited vulnerabilities require immediate fixes, while patches related to AI discoveries can cause instability.

Microsoft fixed 1,413 vulnerabilities in October, surpassing its previous annual record, while CEO Satya Nadella warned that patches related to AI discoveries could destabilize systems. Several outages affected U.S. businesses and agencies, with no evidence of an attack found by authorities.

The FBI Reports Intrusions Using ARTEX

On November 13, the FBI reported the use of ARTEX by cybercriminals and state-sponsored hackers. This open-source artificial intelligence tool was designed to test the cybersecurity resilience of companies. According to the FBI, some intrusions allow attackers to move from an initial target to administrator privileges in less than three hours.

At the New York Times DealBook Summit, Satya Nadella explained that a patch for a critical banking system could previously undergo testing for a week. When a vulnerability is already being exploited, its deployment must now be immediate, he stated.

According to the British Institute for AI Security, open-weight models are only four to seven months behind the most advanced in cyber tasks. These models can be downloaded and run by anyone without protections.

Calif Discovers Vulnerabilities in Java and PHP

On November 17, Calif unveiled Java4Shell, a critical vulnerability discovered using Cyber GPT-6, the OpenAI model. Java powers government agencies as well as about 90% of Fortune 500 companies. Oracle provided a patch on the same day; a federal alert called for its immediate application.

The previous month, the startup had already identified a PHP vulnerability using OpenAI models. This language is involved in the operation of about 70% of websites. Jay Clayton, the Director of National Intelligence, linked the NSA outage to updates made following this discovery.

Thai Duong, CEO of Calif, believes that systems left unpatched will remain exposed to Java4Shell for the next five to ten years.

The Incidents Have Distinct Causes

In less than a week, a global failure affected Microsoft messaging, the FAA suspended all U.S. departures, Chase payments were blocked, Apple experienced malfunctions, and the NSA's internal network became unavailable. In Toronto, a sign indicated the cancellation of U.S. flights following the FAA's decision. Senator Josh Hawley questioned whether it was merely a coincidence. However, intelligence and law enforcement officials testified before Congress that they found no evidence of an attack.

La lettre du soirGratuit · 19 h

Le brief IA que les pros lisent chaque soir

Les 7 actus IA du jour, décryptées en 5 min. Gratuit.

Inclus dès l'inscription : notre sélection des meilleurs guides & comparatifs IA.

Choisissez votre rythme

Gratuit · Pas de spam · Désabonnement en 1 clic

Each incident has a different technical origin: at the FAA, a contractor installed a router; at Microsoft, an anomaly was introduced in July; at Chase, it was a software issue; at Apple, an internal update; and at the NSA, a patch. For Satya Nadella, the Microsoft anomaly was a dormant bug. This week, he also warned that identifying and fixing bugs using AI could cause instability.

Etzioni defends an explanation based on the urgency of updates rather than a coordinated operation: a single attacker would have had to compromise three companies and two federal agencies, disguise their actions as ordinary errors, and evade investigators. He notes that attackers often come forward, whether through a terrorist claim, a ransom demand, or a public statement. The hackers responsible for the MGM casino paralysis in 2023 had contacted authorities a few days after the incident. No claims accompanied these outages.

Chinese hackers from Volt Typhoon maintained access to U.S. critical infrastructure for at least five years before its revelation, positioning themselves to disrupt it during a future crisis. Etzioni does not identify any credible suspects to support the theory of a state warning.

Glasswing Surpasses 10,000 Bugs Discovered

Partners of Glasswing, Anthropic's program aimed at protecting critical software before its offensive use by AI, identified over 10,000 severe or critical bugs during its first month. The code for Windows, iOS, and Chrome is not unlimited; a vulnerability that has been eliminated can no longer be exploited by an attacker.

Angela Gunn, a former Microsoft employee specializing in security, observes that independent researchers are increasingly finding the same vulnerabilities. She sees this as an indication of the limited stock of bugs. For Etzioni, discoveries still outpace the capacity for fixes, but large companies have the financial resources and workforce necessary to catch up.

Etzioni anticipates more bugs, urgent fixes, and outages each month. He also estimates that Windows, iOS, and Chrome could reach their highest level of security in a year or two.

Brief IA — L'actualité IA en français

L'essentiel de l'actualité de l'intelligence artificielle, décrypté et expliqué chaque jour.