Open-source software will replicate your boss and let Claude do their work
No one’s safe.
Image by Cybernews
Follow us
- Open Executive uses Anthropic Claude to mimic a CEO and nine senior executive roles.
- The tool connects to Slack, email, Google Workspace, Notion, Discord, and Telegram.
- Sente Labs says it can approve spending, schedule meetings, direct staff, and set company direction.
- A Cybernews researcher warns broad tool access could expose company documents and employee credentials.
Key Takeaways by nexos.ai, reviewed by Cybernews staff.
Open Executive, an open-source project, uses Anthropic Claude models to provide the knowledge of 9 senior executives and run as a “digital twin” of the CEO.
The new repository on GitHub, which has garnered over 4,800 stars, indicating users’ likes or bookmarks, was first brought to attention by content creator Alberta Devor, known as Alberta Tech.
“If you are worried about your boss replacing you with AI, now you can replace him first,” she said in a sketch video.
Sente Labs, the creators of Open Executive, say the software can be configured as a “digital twin to replicate a busy leader,” and supports spend approval thresholds and integration with corporate knowledge systems, among other things.
In addition, it offers the voice of a team of 9 senior specialists, including a chief financial officer, chief HR officer, and board communications director, which is powered by Anthropic models Claude Sonnet 5 and Claude Opus 5.
Stay updated with our latest stories and follow us on social media
Be the first to discover new stories, ideas, and updates from our team.
Open Executive integrates commonly used collaboration tools, such as Slack, Discord, Telegram, email, Google Workspace, and Notion.
Alex Cohen, who is in charge of change management and transformation at Sente Labs, wrote in an introductory post on LinkedIn that the software was built to operate fully autonomously, capable of directing staff and setting vision and direction.
“It can even act purely as an executive assistant, scheduling meetings and handling basic responses,” Cohen said.
“Sacrificing security for autonomy”
Devor’s video garnered dozens of comments, with users jokingly suggesting that Claude could be a more capable CEO than they currently have, not least because bosses “already use LLMs for everything.”
The release of Open Executive and similar software, however, raises important questions about the risks of allowing AI to handle sensitive data and make decisions.
More than 8 in 10 (82%) of organizations use AI agents, but only 44% have security policies in place to govern them, according to Hostinger data.
Rasa Jurgutytė, an information security researcher at Cybernews, says the project has the potential to “sacrifice security for more autonomy of AI agents.”
Because the agent would have access to multiple company tools and manage its internal documents, it would make it easier for malicious actors to gather large amounts of company information, including employee credentials.
Onboarding a company on the Open Executive app requires sharing information such as the business model, revenue, and strategic priorities, with the option to upload financial documents, as per the project’s GitHub repository.
Creators say data such as the profile YAML, uploaded documents, and the ChromaDB vector store don’t leave users’ local machines or their own volumes in cloud deployments, except for prompts being sent to the Anthropic API.
Sente Labs told Cybernews that because the software is open source, it can be deployed on users’ own infrastructure, such as an on-premises server, “ensuring data never leaves your organizational boundaries.”
In addition, the company says that configurable settings allow users to restrict communication channels to prevent data exfiltration by unauthorized users.
Curious what others think about this story? Contribute your thoughts to the debate below.
Open Executive integrates Sente Labs’ Extensible MCP project, which uses a variety of authorization and logging technologies to “ensure that every OpenExec action is tied back to an actual user request and can be permitted or refused based on the original request.”
“To do so, we integrate with cryptographic signatures and policy languages like Rego and CEL,” the company’s response reads.
Jurgutytė points out that the project is currently in one shared workspace with no per-user data isolation.
This can mean that if, for example, 5 different companies using the same tool could potentially access each other’s data through the tool by accident, it would expose confidential company information.Rasa Jurgutytė, an information security researcher at Cybernews
However, Jurgutytė emphasizes, whether this could happen depends on how authorization is handled.
Sente Labs says that Open Executive is designed for a one-organization deployment, meaning each company runs its own instance on its own infrastructure, so there is no shared workspace to leak across companies.