Addigy, Arcjet Launch AI Controls as Comp AI Raises $34M

9to5Mac reported that Addigy, which builds device management software for managed service providers and IT teams working with Apple, added the Intelligence Suite directly inside its platform rather than as a separate console. The suite includes AI Script Assist, which writes, audits and recommends actions for administrators to approve while following existing role-based access control rules, and natural-language reporting. Jason Dettbarn, Addigy's chief technology officer and founder, said administrators define boundaries and Addigy enforces them.

Addigy said its AI runs entirely inside its own environment and that customer data is never sent to external frontier models. The company also shipped an MCP server that exposes its device management control plane to AI agents, with admin-set guardrails for every action. Prebuilt compliance rules are designed to help Apple IT administrators see, monitor, control and block AI tools across Mac fleets. Nimoy Johnson, an endpoint engineer at Smurfit WestRock, said the reporting capability was among the best he had seen and would help him get more done. Tim Pearson, chief executive and founder of CreativeTechs, said he used the suite to produce a client-facing security risk plan in an hour, compared with one to two business days previously.

SiliconANGLE reported that Arcjet Labs Inc. launched agent runtime security, which tracks what AI agents do once they run in production systems. Security teams receive a list of active agents, and each action can be checked against policy before it goes through; every run is logged for later review. Arcjet said the offering addresses workflows in which agents write to databases, answer support tickets or process customer refunds, often starting from an email or code commit and moving across systems without step-by-step human checks.

Founder and Chief Executive David Mytton said a risky outcome can build over a series of steps that look reasonable on their own. Arcjet links those steps and provides policy controls that act at machine speed, according to the report. Platform teams can send agent activity through existing OpenTelemetry tooling without code changes or another agent on the host, and Claude customers can use Anthropic's Compliance API. Prompts, tool-call parameters and security decisions from separate sessions are stitched into one workflow per agent.

Policies can detect prompt injection or redact personal information before it leaks, and bot detection and rate limits are available. They run on Open Policy Agent and can be changed without redeploying the application, so a team could cap the dollar value of a refund an agent issues without waiting for a release. Each policy check happens before an action executes; Arcjet returns a decision, and the application can halt the call, ask a person to approve it or send the agent an explanation. A second check after the call verifies the result. Native integrations include the Claude Agent SDK, OpenAI Agents SDK, LangChain, Mastra and Microsoft's Agent Framework.

For audits, correlated traces keep actions, inputs and policy evaluations from each execution. Some controls, including personal data detection, run inside the customer's own process, and customers can keep audit records in Arcjet's cloud, a single-tenant or private cloud deployment, or storage they manage themselves. The release moved Arcjet's Python and Go SDKs to version 1.0, after the JavaScript and TypeScript SDK reached that milestone in February and now supports more than 15 frameworks. Arcjet said more than 500 production applications run its software. Its backers include Plural, which led an $8.3 million Series A round last October, as well as Andreessen Horowitz, Seedcamp and former Twilio CTO Ott Kaukver.

SiliconANGLE also reported that Comp AI raised $34 million to advance software that prepares companies for security audits, with the money going toward monitoring that continues after an audit is signed off. The software reads a customer's own systems and documents for organizational context, then drafts policies and risk registers. Agents handle onboarding, gather evidence, watch controls and run vendor security assessments. Dub Technologies Inc. and OpenCode are among more than 1,000 companies using it, according to the report.

The platform's core is open source, and it covers frameworks including SOC 2 and ISO 27001. Its GitHub page positions Comp AI as an alternative to Vanta Inc. and Drata Inc.; Vanta was valued at $4.15 billion in a $150 million round last year. Lewis Carhart, co-founder and chief executive of Comp AI, said the first generation of compliance software proved the process could be automated, while agentic AI lets the company go further. Claudio Fuentes, co-founder and chief operating officer, said compliance had historically been an approximation of security because the work could only be done by hand and on a fixed schedule, and that AI agents remove that limit as attacks get faster.

Comp AI is incorporated as Bubba AI Inc. and runs out of Miami with an office in New York. Carhart started it in January 2025 with Claudio Fuentes and Mariano Fuentes, the chief technology officer. Annual recurring revenue has multiplied 15 times over the past year, according to the company. Roo Capital and Grand Ventures led the Series A round. The pre-seed round, $2.6 million in August 2025, was co-led by Grand Ventures and OSS Capital, with Sentry co-founder David Cramer among the angels. Investors have now put $36.6 million into Comp AI.