Following reports of critical vulnerabilities by LayerX, OpenAI admits that securing AI browsers against indirect prompt injection is an ongoing arms race, raising concerns for user privacy and enterprise adoption.
As OpenAI rolls out new defenses for its agentic browser, the company admits that prompt injection attacks may be an indefinite vulnerability for AI systems.
The UK's National Cyber Security Centre cautions that treating AI prompt injection like SQL injection is a dangerous category error, warning that the vulnerability may never be fully patched.
As 2025 concludes, the Cloud Security Alliance warns that traditional data protection pillars are obsolete against new AI threats like prompt injection and model inversion.