Artificial intelligence has moved well beyond being a simple productivity tool. Today, platforms such as Microsoft Copilot, Zoom AIC, Anthropic’s Claude, and Google Gemini are functioning as active, native participants in the digital workplace — generating outputs, joining meetings, and interacting with both humans and other AI systems in real time.
“The mandate to deploy AI is urgent, but urgency must not lead to excessive risk-taking. Success requires balancing rapid innovation with risk discipline. We enable organizations to execute their AI strategies effectively while unlocking deep insights into their AI investments.”
– Dan Nadir, Chief Product Officer, Theta Lake
According to Theta Lake, this shift has given rise to a new category of communications known as “aiComms” — a term encompassing all human-to-AI and agent-to-agent interactions, and creating an entirely new set of user behaviors. As AI becomes embedded in daily workflows, enterprises face a pressing question: how are employees using it, what are their intentions, and who is governing these conversations?
The risk landscape surrounding aiComms is evolving rapidly and does not fit neatly into traditional compliance frameworks. Unlike conventional workplace communications, AI-generated exchanges tend to be verbose, sprawling, and difficult to investigate retrospectively. Compounding this, the intent behind problematic interactions is not always clear-cut.
Risk professionals must now contend with a spectrum that ranges from deliberate, malicious information gathering to well-intentioned but accidental data exposure. Add to this the risks of prompt manipulation, model hallucination, and the inadvertent leakage of sensitive data, and it becomes apparent why governance specialists are paying closer attention to how AI systems communicate within organisational environments.
Addressing these risks requires more than conventional security tooling. A robust aiComms governance framework begins with data enrichment — supporting existing systems such as Security Information and Event Management (SIEM) platforms by capturing AI conversations in their full context, preserving nuance, intent, and the behavioural signals that only become apparent over time. This over-time monitoring capability is critical, as many emerging risks do not materialise in a single interaction but instead develop across multiple exchanges. Read the full article.
