Enterprise AI deployments now face a critical security gap: preventing sensitive data leaks, compliance violations, and prompt-injection attacks when chatbots, agents, and copilots handle company information. Witness AI has closed a $58 million funding round to deliver what it calls “the confidence layer for enterprise AI,” giving security teams granular control over how employee and agent prompts flow through large-language-model services.
At a Glance
- Witness AI secures $58 million to govern enterprise AI usage.
- The start-up positions itself as the “confidence layer” between companies and AI services.
- The global market for AI security could reach $800 billion to $1.2 trillion by 2031.
- Why it matters: As AI agents begin interacting with one another without human oversight, centralized guardrails become essential to protect data and maintain compliance.
A recent News Of Philadelphia Equity podcast episode unpacked the funding, the market size, and the operational headaches driving demand. Co-host Rebecca Bellan interviewed Barmak Meftah, co-founder and partner at Ballistic Ventures, and Witness AI CEO Rick Caccia to explore what corporations actually fear about AI adoption.
What Enterprises Really Worry About
Caccia told News Of Philadelphia that companies are not simply asking for another security dashboard; they want practical ways to let employees use generative AI while enforcing data-residency rules, role-based access, and audit trails. The concern is twofold:
- Sensitive customer records or source code can travel to third-party model providers.
- Malicious or accidental prompt injections can return harmful or non-compliant outputs.
Meftah added that early adopters initially tried to block AI services outright, only to discover shadow usage and lost productivity. The new approach, he said, is to “enable with guardrails,” a gap Witness AI targets.
The $1 Trillion Market Opportunity
Both guests cited figures from industry models estimating that AI security and governance could swell into an $800 billion to $1.2 trillion annual market by 2031. Drivers include:
- Rapid rollout of chatbots across customer support, finance, and HR
- Proposed U.S. and EU regulations that assign liability for AI-generated harm
- Growing enterprise reliance on API-based large-language-model providers
Caccia argued that traditional data-loss-prevention tools were built for files and email, not dynamic conversational prompts, creating an opening for purpose-built AI controls.
How the Platform Works
Witness AI’s cloud-native service sits inline between corporate networks and models such as OpenAI, Anthropic, or internal Hugging Face instances. According to the company, the platform:
- Intercepts every prompt before it reaches the model
- Applies encryption or redaction based on classification tags
- Logs interactions for compliance teams in searchable form
- Enforces rate limits and content policies to block injection attempts
Meftah said Ballistic Ventures led the $58 million round because the architecture can scale without adding milliseconds of latency, a deal-breaker for real-time customer applications.
When AI Agents Talk to Each Other
A key inflection point, Caccia noted, is the emerging scenario where multiple AI agents exchange data autonomously, such as a supply-chain bot negotiating with a logistics bot. Without oversight:
- Confidential pricing data can leak outside corporate boundaries
- Compliance rules may be bypassed because no human reviews the conversation
- Attack surfaces multiply if one compromised agent issues malicious instructions
Witness AI claims its policy engine can insert itself into agent-to-agent traffic, effectively brokering trust between bots the same way it governs human prompts.
Competitive Landscape
The start-up joins a crowded field that includes:
- Model vendors adding native safety filters
- Cloud providers embedding AI security into managed services
- Niche start-ups focusing on prompt firewalls or bias detection
Caccia argued that neutrality gives Witness AI an edge: “We don’t build models, we don’t run clouds, so customers see us as Switzerland,” he told News Of Philadelphia.
Funding Details and Hiring Plans
Although the company declined to disclose valuation, Olivia Bennett Harris reported that the round was “significantly oversubscribed” in a matter of weeks, signaling strong investor appetite for AI tooling. Proceeds will fund:
- Engineering hires across data encryption and low-latency networking
- Sales teams focused on Fortune 500 financial and healthcare accounts
- Product development for on-prem deployment options demanded by regulators
Meftah confirmed Ballistic co-led the investment alongside a syndicate of unnamed returning backers.
What Comes Next
Witness AI plans public reference customers by the fourth quarter and hinted at partnerships with major cloud hyperscalers. Caccia told News Of Philadelphia that early adopters are already stress-testing the platform against red-team prompt attacks, with results feeding quarterly model updates.
Listeners can catch the full conversation on the News Of Philadelphia Equity podcast, available on YouTube, Apple Podcasts, Overcast, Spotify, and other platforms. The show also streams updates on X and Threads under the handle @EquityPod.
