BlackFog has launched ADX Vision 2.0, extending its AI security product to cover data sent by employees and AI agents to external AI services.
The update adds seven layers of prompt protection, custom controls for sensitive data, optional prompt auditing, and tools to route requests to approved large language model services instead of personal accounts.
The new controls apply to both prompts written directly by staff and requests generated by AI agents acting on their behalf, allowing organisations to apply the same data rules across human and automated activity.
The release comes as businesses try to manage broader use of generative AI tools in the workplace. Research cited by BlackFog found that 49% of surveyed employees had used AI tools not approved by their employer, with some sharing employee information, financial statements, and sales data.
Prompt controls
The seven protection layers are designed to identify prompt injection attempts and other manipulated or malicious prompts. They include context switching protection, obfuscation protection, jailbreak detection, data exfiltration protection, dangerous prompt pattern detection, fuzzy keyword detection, and structural anomaly detection.
Organisations can also apply company-wide or departmental controls to identify specific forms of sensitive information, including credit card numbers, Social Security numbers, and sector-specific terminology.
The update also allows AI requests to be redirected to sanctioned corporate AI services. BlackFog gave the example of an employee opening a personal ChatGPT account and being routed instead to the organisation's enterprise AI instance.
The platform introduces optional prompt auditing for what BlackFog described as leading frontier large language models, giving organisations more visibility into what information is being shared with AI systems.
Usage tracking
ADX Vision 2.0 also includes a unified view of token consumption across large language models, allowing organisations to see where tokens are being used and by whom, while helping identify possible token theft through stolen API keys or credentials.
BlackFog positions the product as an endpoint-based control layer that operates before data reaches an AI service, giving organisations oversight across AI interactions regardless of which tool or model is used.
Dr Darren Williams, Founder and Chief Executive Officer of BlackFog, said the shift from employee use of AI tools to more autonomous agents had changed the nature of the security problem.
"AI is quickly moving from something employees interact with to something that can make decisions and take action on their behalf. That changes the security equation. When an agent is sending prompts on an employee's behalf, there's no human in the loop to catch a manipulated instruction or a leak of sensitive data. ADX Vision 2.0 inspects agentic prompts with the same protections as those written by humans," Williams said.
The latest release builds on earlier shadow AI discovery and control functions. BlackFog is now positioning the software as a broader governance layer as organisations deploy more autonomous AI systems.
BlackFog was founded in 2015 and focuses on anti-data exfiltration software. Its platform is used by more than 500 enterprises, government agencies, and critical infrastructure operators worldwide.
ADX Vision 2.0 is available for Windows and macOS.