AI Agents Are Getting Socially Engineered, Says AI Threat Report 2026. OrcaRouter Launches Free Firewall and Guardrails

New report warns that prompt injection is becoming the phishing attack of the AI era. OrcaRouter is making AI agent security free for all users.

SAN FRANCISCO, June 18, 2026 /PRNewswire/ -- AI agents are becoming the next target of social engineering, according to the AI Threat Report 2026 released today by OrcaRouter Security Research.

The report argues that prompt injection is emerging as the phishing attack of the AI era. Instead of tricking people into clicking malicious links, attackers increasingly target AI systems directly through emails, documents, websites, and tool outputs that agents read and act upon.

"In the last decade, attackers learned how to socially engineer people. In this decade, they're learning how to socially engineer agents."

The warning comes after a year of high-profile AI security incidents, including prompt injection attacks, agent hijacking, data exfiltration, and denial-of-wallet attacks that can generate massive AI bills without stealing a single record.

As AI agents gain access to company knowledge bases, email systems, source code, CRMs, and external tools, they are becoming one of the most powerful—and least protected—systems inside modern organizations.

To help organizations defend against this new class of threats, OrcaRouter today launched Firewall and Guardrails, available at no additional cost for all OrcaRouter users.

Guardrails screens prompts and model outputs for prompt injection, jailbreaks, sensitive data leaks, and policy violations.

Firewall governs what agents can do, controlling tool calls, MCP actions, network access, approvals, and spending limits.

Together, they help organizations control both what AI agents see and what AI agents do.

Unlike traditional security products that sit outside the AI stack, Firewall and Guardrails operate directly inside the OrcaRouter AI gateway and can be enabled without changing application code.

The AI Threat Report 2026 concludes that AI security can no longer rely on protecting users alone.

Organizations must start protecting the agents acting on their behalf.

Firewall and Guardrails are available today for all OrcaRouter users at no additional cost.

Resources

AI Threat Report 2026
https://docs.orcarouter.ai/whitepapers/ai-threat-report-2026

OrcaRouter Firewall
https://docs.orcarouter.ai/security/firewall/overview

OrcaRouter Guardrails
https://docs.orcarouter.ai/security/guardrails/overview

OrcaRouter
https://www.orcarouter.ai

Media Contact
OrcaRouter
+1 6506097501
[email protected]

SOURCE OrcaRouter