What happened
On June 22, 2026, OrcaRouter (OpenAI-compatible LLM gateway, Singapore) published its AI Threat Report 2026 documenting 14 key risks across four threat categories — noting prompt injection rose 340% YoY (OWASP Q1 2026), average successful attack completes in 42 seconds, and 13% of organizations have been breached via an AI model. Simultaneously made its agent Firewall and input/output Guardrails free to all users attached to an existing API key, with no separate integration required.
Why it matters
Removing the cost barrier for gateway-level LLM controls (firewall + I/O guardrails) broadens adoption of baseline AI security controls to startups and smaller teams that previously couldn't justify the spend — directly relevant as prompt injection scales.
Applicability
Any team operating LLM-based applications or agents in production should attach OrcaRouter's free Firewall and Guardrails to their existing API keys as a baseline control; security architects should review the AI Threat Report 2026 for the 14-risk taxonomy.