The AI Blind Spot Is Growing Every Day

Your employees are already using AI tools to write emails, summarise documents, generate code, and analyse data. Most of them aren't telling you. The question isn't whether AI is being used in your organisation. It's whether you know what data is going into it.

36%

of employees upload sensitive company data to AI tools

Strategic plans, technical data and financials are the most common uploads, with 24% also putting customer PII into public AI platforms.

USD $4.63M

global average cost of a shadow AI breach

USD $670K more than standard breaches. 1 in 5 organisations already affected

77%

of GenAI users paste data into AI tools

82% of that pasting happens from unmanaged personal accounts, and 22% of pastes contain PII or PCI data.

AI Is the Biggest Blind Spot in Your Organisation

Data Leakage

Employees paste source code, client data, financial models, and legal documents into AI chatbots. Once it’s in the prompt, it’s in the AI provider’s system.

Shadow AI

Your team is using AI tools you haven’t approved, on accounts you don’t control, with data you can’t see. Standard endpoint security tools don’t detect AI usage.

Compliance Risk

Regulators worldwide are scrutinising AI data handling: Australian privacy authorities, Singapore’s PDPC, and the EU AI Act all expect demonstrable oversight. “We didn’t know” is not a defence when client data appears in AI prompts.

How ActivityPulse Detects AI Usage

Every interaction with AI tools is captured in real time. Prompts are analysed for sentiment and classified for risk. Captured prompt data is the most sensitive data ActivityPulse holds, and it stays under your control. Hosted in your region: Australian hosting as standard, with in-region hosting available for APAC customers.

Complete AI Usage Visibility Across the Major AI Services

From mainstream consumer tools to AI services hosted in jurisdictions your data policies may restrict.

Major Consumer AI

ChatGPT, Claude, Google Gemini, Microsoft Copilot, DeepSeek, Perplexity, Grok, Mistral, Meta AI

Popular Alternatives

Poe, HuggingChat, You.com, Phind, Pi, Character.AI, Writer, Jasper, Komo, Andi

Workplace AI

Microsoft 365 Copilot, Google AI Studio, NotebookLM, Gemini Enterprise, Google Flow, Opal

Chinese-Origin AI (Jurisdiction and Data Residency)

Baidu ERNIE, Alibaba Qwen, ByteDance Doubao, Moonshot Kimi, Zhipu ChatGLM, MiniMax Hailuo

Russian-Origin AI

Sber GigaChat, Yandex GPT

Always Growing

New AI services can be added via server configuration. No agent update required.

How Data Flows to AI. All Captured.

Typed Prompts

When an employee types a question or instruction into an AI service, ActivityPulse captures the text. Detects submission via Enter key, typing timeout, or app switching.

Pasted Content

When an employee pastes text into an AI service, often source code, email content, or document text, it’s captured as a separate event. This is frequently the highest-risk data flow.

Copied Responses

When an employee copies text from an AI response, such as code snippets, generated content, or summaries, it’s captured. This shows data flowing out of AI services back into your organisation.

How We Govern Our Own AI Prompt Capture

Private by default. Powerful by choice. AI prompt capture is a sensitive module: it ships off by default and is enabled deliberately, per agent, under your policy. Input activity is counted as keystrokes and clicks, not the keys typed, and the camera and microphone module records which apps used them, not the audio or video. Every sentiment and risk score carries a methodology panel showing how it was calculated, so the intelligence you act on can be explained to your board, your employees and your regulator.

πŸ‡ͺπŸ‡Ί EU AI Act (2024)

The world’s first comprehensive AI law. Requires organisations to maintain records of AI system usage, conduct risk assessments, and demonstrate transparency. Fines up to 7% of global revenue.

πŸ‡¦πŸ‡Ί Australia’s National AI Plan (2025)

Existing privacy and consumer laws apply to AI. The OAIC requires disclosure of automated decision-making by December 2026. A new AI Safety Institute (operational 2026) will assess risks. The Privacy Act reforms make AI data governance a board-level obligation.

🏒 ISO/IEC 42001 (2023)

The first international standard for AI management systems. Requires organisations to establish, implement, and continuously improve AI governance. Auditors will ask for evidence.

πŸ”’ Industry Mandates

Finance (APRA CPS 230/234), healthcare (My Health Records Act), legal (professional conduct rules), and Australian government agencies (DTA AI Policy v2.0, effective December 2025) all require demonstrable AI data controls.

πŸ‡ΈπŸ‡¬ Singapore: PDPC and the Model AI Governance Framework

Singapore’s PDPC applies the PDPA to personal data used in AI systems, with advisory guidelines for AI recommendation and decision systems issued in 2024. The Model AI Governance Framework, extended to generative AI in 2024, sets the regional benchmark for demonstrable AI oversight. Organisations are expected to show how AI use of data is overseen and controlled.

Tue 2:14 PM
Opened ChatGPT in Chrome
Tue 2:15 PM
Pasted 847 lines of proprietary source code
Tue 2:16 PM
Prompt: “refactor this to use async/await”
Tue 2:18 PM
Copied AI-generated code response (312 lines)
Tue 2:22 PM
Uploaded client-report-Q1-2026.xlsx to Claude
Tue 2:23 PM
Prompt: “summarise the key findings in this report”
Tue 2:31 PM
Opened DeepSeek (unapproved, Chinese-origin AI)
Tue 2:32 PM
Pasted internal API credentials into prompt

The Investigation Timeline traces every AI interaction in chronological order. Filter by service, prompt type, or sentiment. Export as branded PDF or DOCX for compliance records.

Build Your AI Policy with Data, Not Guesswork

Which AI services are employees using?

How frequently are they using them?

What type of data are they sharing?

Are they using approved tools or shadow AI?

Is sensitive data appearing in prompts?

Which departments are heaviest AI users?

Are files being uploaded to AI services?

Are employees using foreign-origin AI tools?