Enterprise security teams gain a new layer of control over agentic AI with ADX Vision 2.0, an update that inspects prompts, governs which AI services employees and agents can reach, and tracks token use across the organization. The release responds to a fast-moving shift inside companies: AI is no longer just a chat tool that employees type into, it is a set of autonomous agents that send prompts and act on corporate data without a human reviewing each step.<\/p>\n
ADX Vision 2.0 operates as an independent control layer on the endpoint, inspecting prompts before they reach any AI service. The goal is to give organizations oversight and an audit trail across every AI interaction, whether a prompt comes from an employee or from an AI agent acting on their behalf. The platform decides what an AI is allowed to do with data, regardless of which tool or model is in use.<\/p>\n
The update builds on the earlier Shadow AI discovery and control features and extends them into a broader security and governance layer for organizations deploying increasingly autonomous AI systems.<\/p>\n
ADX Vision 2.0 introduces seven layers of protection aimed at identifying prompt injection attempts and other malicious or manipulated prompts:<\/p>\n
The same protections apply whether a prompt is written by an employee or generated by an AI agent on their behalf. That coverage extends to autonomous workflows where there is no human oversight of what is being sent. Custom controls can also be applied at a company or departmental level to flag sensitive information, including credit card numbers, Social Security numbers, and domain-specific terminology.<\/p>\n
Beyond prompt inspection, ADX Vision 2.0 adds three operational capabilities for security and finance teams.<\/p>\n
Redirecting to sanctioned AI services.<\/strong> Organizations can route AI requests to officially licensed LLM services, which enforces AI policy without blocking employees from using the technology. In practice, an employee who opens a personal ChatGPT account can be automatically redirected to the company’s enterprise AI instance, keeping corporate data inside approved environments.<\/p>\n Prompt auditing for frontier LLMs.<\/strong> An optional prompt auditing feature supports governance and compliance requirements, including a record of what is being shared with AI systems as regulation evolves.<\/p>\n Unified token consumption view.<\/strong> A single dashboard shows where tokens are being consumed and by whom, supporting financial planning and cost efficiency. The same view helps spot potential token theft, where attackers use stolen API keys or credentials to run up usage at the organization’s expense.<\/p>\n When an agent sends prompts on a user’s behalf, there is no human in the loop to catch a manipulated instruction or a leak of sensitive data. Inspecting prompts on the endpoint, before they leave the device, means the control applies to the prompt itself rather than to a particular AI service or model. That distinction matters as companies mix sanctioned enterprise instances, personal accounts, and third-party agents on the same network.<\/p>\n ADX Vision 2.0 is an enterprise AI security and governance platform that inspects prompts on the endpoint before they reach any AI service, applies policy controls, and audits AI interactions across the organization.<\/p>\n The seven layers are context switching protection, obfuscation protection, jailbreak detection, data exfiltration protection, dangerous prompt pattern detection, fuzzy keyword detection, and structural anomaly detection.<\/p>\n The platform provides a unified view of token consumption across LLMs to support cost planning and to flag potential token theft. It can also redirect requests from personal AI accounts to an organization’s sanctioned enterprise AI instance.<\/p>\nWhy Endpoint-Level Inspection Matters<\/h2>\n
FAQ<\/h2>\n
What is BlackFog ADX Vision 2.0?<\/h3>\n
What are the seven layers of prompt protection in ADX Vision 2.0?<\/h3>\n
How does ADX Vision 2.0 handle tokens and personal AI accounts?<\/h3>\n