Large language models (LLMs) and AI agents are becoming deeply integrated into web browsers, search engines and automated content-processing pipelines. While these integrations can expand functionality, they also introduce a new and largely underexplored attack surface.
One particularly concerning class of threats is indirect prompt injection (IDPI), in which adversaries embed hidden or manipulated instructions within website content that is later ingested by an LLM. This article shares in-the-wild observations from our telemetry, including our first observed case of AI-based ad review evasion.
Read more…
Source: Palo Alto Unit 42
Sign up for the Cyber Security Review Newsletter
The latest cyber security news and insights delivered right to your inbox
Related:
- Another BRICKSTORM: Stealthy Backdoor Enabling Espionage into Tech and Legal Sectors
September 24, 2025
Google Threat Intelligence Group (GTIG) is tracking BRICKSTORM malware activity, which is being used to maintain persistent access to victim organizations in the United States. Since March 2025, Mandiant Consulting has responded to intrusions across a range of industry verticals, most notably legal services, Software as a Service (SaaS) providers, Business Process Outsourcers (BPOs), and Technology. ...
- US federal agency breached by hackers using GeoServer exploit
September 24, 2025
In mid-July 2024, a threat actor managed to break into a US Federal Civilian Executive Branch (FCEB) agency by exploiting a critical remote code execution (RCE) vulnerability in GeoServer, the government has confirmed. In an in-depth report detailing the incident, the US Cybersecurity and Infrastructure Security Agency (CISA) outlined how the attackers leveraged CVE-2024-36401, a 9.8/10 ...
- UK: Man arrested in connection with cyber-attack on airports
September 24, 2025
A person has been arrested in connection with a cyber-attack which has caused days of disruption at several European airports including Heathrow. The National Crime Agency (NCA) said a man in his forties was arrested in West Sussex “as part of an investigation into a cyber incident impacting Collins Aerospace”. There have been hundreds of flight ...
- Top auto insurance firm leaked over 5 million records
September 24, 2025
ClaimPix, a company which streamlines car insurance claims, was leaking sensitive customer data on the clearweb, including people’s phone numbers, and email addresses, an expert has warned. Security researcher Jeremiah Fowler, known for hunting down misconfigured and unprotected databases, recently found one such instance containing 5.1 million files, sharing his findings with WebsitePlanet. The archive was ...
- Serious Microsoft Entra flaw could have let hackers infiltrate any user – patch now
September 22, 2025
Security researchers have found a critical vulnerability in Microsoft Entra ID which could have allowed threat actors to gain Global Administrator access to virtually anyone’s tenant – without being detected in any way. The vulnerability consists of two things – a legacy service called “actor tokens”, and a critical Elevation of Privilege bug tracked as CVE-2025-55241. ...
- Stellantis detects breach at third-party provider for North American customers
September 22, 2025
Stellantis detected unauthorized access to a third-party service provider’s platform that supports its North American customer service operations, the company said in a statement on Sunday. The automaker said the incident, which is under investigation, exposed only basic contact information and did not involve financial details or sensitive personal data. Stellantis did not specify how many ...