The New AI Guardrails: Security Solution or Surveillance Tool?

OPINION: The C-suite is currently in a state of panic over 'security headaches' brought on by the rapid adoption of AI. But while executives fret over data leaks, there is a more insidious risk unfolding: the infrastructure being built to 'secure' AI will likely be weaponized to disproportionately police and surveil the lowest-paid workers in the building.

As TechCrunch first reported, the startup Pluto Security is leading the charge in this new era of oversight. Pluto recently emerged from stealth with $42 million in funding—comprising an $8.5 million seed round led by TLV Partners and a $34 million Series A led by Jump Capital. Other investors include the founders of cybersecurity firm Armis, Ansa Capital, and Skinos, the venture firm of Check Point co-founder Shlomo Kramer.

Pluto CEO and co-founder Shahar Bahat tells TechCrunch that the risk is driven by employees using AI in 'unexpected ways,' noting that new AI tools can appear in a company's environment within two hours of release. Bahat highlights a specific trend of 'vibe-coded' apps created by non-developers—specifically noting that finance departments are the primary source of these risks, followed by marketing and legal teams. In one instance, Pluto discovered a vibe-coded app created by a finance employee at a Fortune 500 company to manage churn and customer information that had been left exposed to the internet.

While Bahat frames this as a technical necessity, the mechanisms Pluto uses are a blueprint for total employee surveillance. The software doesn't just monitor tools; it tracks 'agent behavior' in real time. To achieve this, Pluto connects to device management and endpoint security systems to deploy sensors. These sensors gather a granular trail of data, including prompts, commands, running processes, permissions, and a history of the user's previous activity.

Most concerning is the ability to intervene. Bahat explained to TechCrunch that Pluto can 'inject context' back into a session to change how an agent works or stop an action if an agent attempts to gain permissions beyond what was intended.

Pluto is already scaling this surveillance apparatus, reporting over $2 million in annual recurring revenue and more than 30 enterprise customers, including AlphaSense, Sixt, Workato, and Oscar Health. They have also launched 'Plutonium,' a free platform for scanning AI skills and connectors to draw in security professionals.

When we move from monitoring 'user behavior' to 'agent behavior,' as Bahat puts it, we aren't just securing a network. We are creating a high-resolution map of how every single employee thinks and works. In a corporate hierarchy, these 'security policies' are rarely applied equally; they are almost always used to tighten the leash on those at the bottom.