Kihan Debuts Constitutional Computing to Enforce AI Agent Authority Before Execution

Kihan announced its formation and launched Constitutional Computing, an architecture designed to make AI governance enforceable at the point of execution, on November 5, 2026. The San Francisco company said the system separates an AI agent's capability from its permission to act. EIN Presswire published the announcement.

How Kihan's Authorization Architecture Works

Constitutional Computing takes its name from governance rather than model training. Authority is written down, signed and enforced outside the AI, according to the company. Each consequential action is evaluated against governing authority before execution, with a cryptographically verifiable permit bound to the specific operation and its parameters. The executing tool verifies that permit before acting, and on protected execution paths no valid permit means no execution. Tamper-evident records show what was authorized, refused or referred for human approval.

Kihan cited research published by the Cloud Security Alliance in April 2026 finding that 53% of surveyed organizations reported AI agents exceeding their intended permissions, while 47% experienced an AI-agent security incident in the preceding year.

"AI agents are becoming extraordinarily capable, but no system, however capable, should be trusted to enforce its own limits," said Jonathan Hirshon, co-founder and CEO of Kihan. "Telling an AI to behave itself isn't security. Authorization has to be independent of the AI, verifiable and enforceable before anything consequential happens."

Kihan's first product, Inktomi, is being developed to bring Constitutional Computing to enterprise AI environments, including agents built on models from Anthropic, OpenAI and Google, agent frameworks such as Hermes Agent, and tool ecosystems using the Model Context Protocol. Its decision engine is already operational and evaluates specific actions against signed rules, determining whether to permit, refuse or escalate a request while preserving a verifiable decision record. Production enforcement integrations are under development. Inktomi is designed to run on customers' own infrastructure and complements existing identity management, cybersecurity and AI governance technologies, the company said. Kihan is initially targeting financial services, insurance, healthcare, legal services and other regulated industries, and is exploring integration with the NVIDIA OpenShell runtime and the IBM watsonx.governance platform. A live demonstration of Inktomi can be arranged through Kihan's website.

CrowdStrike, NVIDIA and Impart Ship Agent Controls

CrowdStrike announced the AI Kill Switch in Falcon Guardian on November 3, 2026, giving security teams the ability to terminate rogue AI agents, block relaunch and cut network access, TradingView reported. NVIDIA introduced the Open Agent Safety Platform on October 30, 2026, combining open-source software with hardware-based monitoring, including OpenShell for sandboxing and Sentry, which runs on BlueField-4 units to quarantine agents in milliseconds, Forbes reported. Impart Security launched its Runtime Defense Harness on November 4, 2026, letting security teams deploy AI-written virtual patches within 60 minutes of a vulnerability disclosure, FinancialContent Business Page reported.

IEEE Computer Feature and Inktomi Evaluations

The research underlying Kihan's architecture has been accepted for publication as a two-part feature in IEEE Computer. The first article, "Checking LLM Outputs Before They Become State: The Problem and the Admission Gate," is scheduled for the magazine's December 2026 special issue on AI Governance, followed in January 2027 by "Checking LLM Outputs Before They Become State: Prompt Framing Across Four Models." Initial enterprise evaluations of Inktomi are targeted for December 2026.