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:
- Critical Buffer Overflow in Palo Alto Networks PAN-OS User-ID Authentication Portal (CVE-2026-0300)
May 6, 2026
On May 6, 2026, Palo Alto Networks published a security advisory for CVE-2026-0300, a critical unauthenticated buffer overflow vulnerability affecting PAN-OS PA-Series and VM-Series firewall appliances. Prisma Access, Cloud NGFW, and Panorama appliances are not affected by this vulnerability. The vulnerability carries a CVSSv4 score of 9.3 and has been confirmed as exploited in the wild ...
- Update WhatsApp now: Two new flaws could expose you to malicious files
May 5, 2026
Meta has published a new security advisory for messaging app WhatsApp, announcing patches for two vulnerabilities. WhatsApp has fixed two security flaws that could be abused to interfere with how media and attachments are handled on your device. There is no evidence that either bug has been exploited in the wild. These bugs don’t automatically infect ...
- DAEMON Tools software compromised with a malicious payload
May 5, 2026
In early May 2026, Kaspersky researchers identified installers of the DAEMON Tools software, used for mounting disk images, to be compromised with a malicious payload. These installers are distributed from the legitimate website of DAEMON Tools and are signed with digital certificates belonging to DAEMON Tools developers. Kaspersky analysis revealed that the software installers have been ...
- Trellix confirms data breach after hack of ‘a portion’ of its source code
May 5, 2026
Cybersecurity giant Trellix has confirmed suffering a cyberattack in which threat actors accessed parts of its source code. In a brief announcement published on its website, Trellix said it had identified “unauthorized access to a portion of source code repository”. As soon as it spotted the intrusion, the company brought in third-party security experts to ...
- Quasar Linux (QLNX) – Inside a Full-Featured Linux RAT
May 4, 2026
In previous research, Trend Micro have demonstrated how AI can be used to improve detection accuracy when new malware families emerge, particularly those that reuse or share code from open-source repositories. In this blog entry, Trends Micro researchers present another compelling finding from the same approach. Trend Micro platform recently flagged an unusual Linux implant with ...
- Thousands of Facebook accounts stolen by phishing emails sent through Google
May 4, 2026
Researchers have uncovered a long-running phishing operation that abuses trusted Google services to hijack tens of thousands of Facebook accounts. The compromised Facebook accounts are mainly business and advertiser profiles, which criminals can monetize after gaining access and control. The attackers found a way to send phishing emails that come “through Google,” making them look legitimate ...