In the race to gain a competitive edge, organizations are increasingly training artificial intelligence (AI) models on sensitive data. But what if a seemingly harmless AI model became a gateway for attackers?
A malicious actor could upload a poisoned model to a public repository, and without realizing it, your team could deploy it in your environment. Once active, that model could exfiltrate your sensitive machine learning (ML) models and fine-tuned large language model (LLM) adapters. With access to these adapters, attackers could replicate your custom tuning and optimizations, exposing sensitive information embedded in fine-tuning patterns. Palo Alto Networks researchers recently uncovered two vulnerabilities in Google’s Vertex AI platform. These vulnerabilities could have allowed attackers to escalate privileges and exfiltrate models. We
Read more…
Source: Palo Alto Unit 42
Related:
- Hasta la vista, Hastalamuerte: An Overview of The Gentlemen’s TTPs
March 19, 2026
In face of so many new ransomware brands, and still remaining RaaS operations such as Medusa, Qilin, and DragonForce, prioritizing is not an easy task to accomplish. However, despite the amount of groups conducting attacks for extortion, the TTPs do not change that much; unless we are talking about Cl0p, Akira and other groups that ...
- CISA urges companies to secure Microsoft Intune systems after hackers mass-wipe Stryker devices
March 19, 2026
The U.S. Cybersecurity and Infrastructure Security Agency (CISA) has warned companies to secure systems for managing their fleets of employee devices after pro-Iran hackers broke into medical tech giant Stryker and mass-wiped thousands of its phones, tablets, and computers. The agency said on Thursday that it was urging companies to take action and confirmed it was ...
- DarkSword: Second iOS exploit chain in a month targeting iPhone users
March 18, 2026
A new exploit kit targeting iPhone users and stealing their sensitive data is being abused by “multiple” spyware vendors and suspected nation-state goons, security researchers said on Wednesday. The exploit kit, called DarkSword, has been in use since at least November 2025. It supports iOS versions 18.4 through 18.7, and exploits six different vulnerabilities to deploy ...
- Over 29 million secrets were leaked on GitHub in 2025, and AI really isn’t helping
March 18, 2026
Vibe-coding may seem great for quickly shipping products, but inexperienced developers are leaving gaping cybersecurity holes that are causing breaches and exposures left and right. This is according to GitGuardian’s latest report, the “State of Secrets Sprawl” paper that was just released. In the research document, the organization said 2025 was the year when AI adoption ...
- Unpacking a new Horabot campaign in Mexico
March 18, 2026
In this instalment of Kaspersky SOC Files series, Kaspersky researchers will walk you through a targeted campaign that our MDR team identified and hunted down a few months ago. It involves a threat known as Horabot, a bundle consisting of an infamous banking Trojan, an email spreader, and a notably complex attack chain. Although previous research ...
- Apple patches WebKit bug that could let sites access your data
March 18, 2026
WebKit vulnerabilities refer to security flaws in Apple’s web rendering engine, which powers Safari, Mail, and the App Store on iOS and macOS. What this means is that the CVE-2026-20643 vulnerability makes it possible for a malicious website to pretend to be another site, maybe one you trust, and then read or steal information that should ...