ThreatAware has launched Dimension, a workspace for building and running custom security applications with governance controls built in. The useful promise is not simply faster app-making: teams can set which data and actions each app may access, and require approval for changes to live systems.
Watch Desk analysis
What happened
Dimension has three components: Data connects to an organisation’s existing systems, Studio is where applications are built, and Actions lets them make controlled changes to live systems. Applications can be created from natural-language descriptions using AI Composer, built with a coding agent such as Claude Code through MCP, or adapted from an App Library template.
The announcement says each app can access only the data feeds and actions connected to it. Actions such as disabling a user or changing cloud configuration must be explicitly configured and approved, and are recorded in an audit log. The announcement and product details also describe shared workspaces, permissions, version control and isolated tenancy.
Why it matters
ThreatAware says 77% of surveyed security decision-makers had used AI coding assistants to build internal security tools in the previous 12 months. It also says 45% reported that self-built tools did not receive the same review and approval as other software, while 41% said they had no formal policy, or did not enforce one, for AI-built security applications. Those figures come from research cited in the product announcement, not an independent assessment here.
The tension is familiar: teams can now make bespoke tools quickly, but a tool that touches sensitive data or changes live systems needs more than a clever prompt and a favourable name. Dimension’s answer is to bundle development with access controls and approval processes.
Our read
This is a relevant attempt to make governance part of the build process rather than a meeting someone remembers to schedule later. The practical test will be whether teams can use the controls without finding ways around them, and whether the system’s permissions and audit trail stand up to real security work. For organisations weighing it, the details that matter are what it can connect to, who approves actions, and how those actions are recorded.
What to watch
- Which integrations and application templates are available to customers.
- How approvals, permissions and audit records work in day-to-day use.
- Whether teams adopt the workspace, and whether it reduces unmanaged or delayed internal tools.
Discussion spark: For internal security tools built with AI, should every action affecting live systems require human approval, or can tightly limited permissions and audit logs be enough?
Sources and evidence
- Source update (27 October 2026, 12:13 UTC)
- Source update (27 October 2026, 12:13 UTC)
Watch Desk is operated by WittyWires as an independent cross-cutting AI news tracker. It does not speak for the organisations or people it covers.