After facing criticism over its open and unpredictable AI tools, Quill is taking a different approach. The startup is emphasizing security and user control while building what it calls an “agentic AI” that helps users manage their workflows. This new direction aims to balance powerful AI capabilities with privacy and safety, setting it apart from earlier, more risky tools like OpenClaw.

Rethinking AI Assistants with Privacy in Mind

Many users are eager to use AI tools that act more like assistants than simple transcription bots. But past experiences have shown that rushing these tools out can lead to problems. OpenClaw, for example, was seen as too unpredictable and risky, with some fearing it could make harmful decisions or mishandle data. Quill wants to avoid these pitfalls by designing its AI with security at the core.

Instead of just logging meetings or copying conversations, Quill’s AI—called Quilliam—integrates directly into collaboration platforms like Slack, Notion, Salesforce, and more. It gathers context from these tools and helps users take action based on what it learns. The company stresses that its system is built with privacy-first policies, meaning data stays local whenever possible, and users have full control over what the AI can access or do.

Built for Context and Control

Quill estimates that workers spend most of their day in meetings, calls, and messaging, but often struggle to make sense of all the information recorded. To address this, Quilliam connects to many tools through a protocol called Model Context Protocol (MCP). This allows the AI to remember ongoing conversations and past interactions, enabling it to offer relevant suggestions, automate workflows, and help generate documents like emails or reports.

Michael Daugherty, Quill’s founder and CEO, explains that AI is most useful when it has more context about a person’s goals and contacts. Instead of providing generic summaries, Quill aims to deliver tailored insights. That way, the AI can support individual workflows more effectively, without replacing human judgment.

Security remains a top priority. Quill’s design is “local-first,” meaning transcription and voice recognition happen on the user’s device. Audio data never leaves the device unless explicitly authorized. Enterprises can set up configurable endpoints, ensuring they retain control over their data at all times. For example, a research agent might develop a plan and request access to specific tools. The user can approve or deny each request, and once the task is done, access is revoked automatically.

Secure, Flexible, and User-Centric AI

Unlike OpenClaw, which was more open and could potentially take unpredictable actions, Quill emphasizes transparency and human oversight. The company states that users can supervise the AI’s steps, approve plans before execution, and limit the tools it can access. This approach minimizes the risk of unintended consequences, making the AI a collaborative assistant rather than an autonomous actor.

Quill offers flexibility for different environments. Its inference models can run in the cloud through partners like Google Vertex or AWS Bedrock, or locally on the user’s hardware for maximum isolation. This setup is ideal for organizations with strict privacy requirements or sensitive data.

Overall, Quill’s focus is on creating a trustworthy AI that enhances productivity without compromising security. By prioritizing user control and local data handling, the startup hopes to rebuild confidence in AI assistants and demonstrate that security and innovation can go hand in hand.

Inspired by

• https://www.computerworld.com/article/4137260/after-openclaw-backlash-quill-bets-on-security-by-design-agentic-ai.html

Sources

• csoonline.com
• csoonline.com