Pluto Security Raises $42M to Combat Shadow AI and Agent Risks
Enterprise security startup emerges from stealth as non-developer vibe-coding creates widespread system vulnerabilities.
Pluto Security has emerged from stealth with $42 million in total funding to address critical enterprise risks stemming from unmonitored AI tools and autonomous agents. Led by an $8.5 million seed round from TLV Partners and a $34 million Series A led by Jump Capital, Pluto Security targets the rapid proliferation of non-developer "vibe-coded" applications across corporate departments. As employees deploy unvetted AI components within hours of release, Pluto Security equips Chief Information Security Officers (CISOs) with real-time visibility and active execution controls to contain rogue agents and prevent data exfiltration.
Key Details
The investment reflects growing anxiety among enterprise security leaders as employee adoption of artificial intelligence outpaces traditional security governance. Enterprise workers increasingly construct custom workflows using lightweight LLMs, browser extensions, and Model Context Protocol (MCP) servers without IT oversight.
- Funding Milestones: Pluto Security secured an $8.5 million seed round led by TLV Partners with participation from Armis founders, followed by a $34 million Series A led by Jump Capital, alongside Skinos and Ansa Capital.
- Commercial Traction: Since launching commercial sales in April 2026, Pluto Security surpassed $2 million in Annual Recurring Revenue (ARR) across more than 30 enterprise clients, including Oscar Health, Workato, Sixt, and AlphaSense.
- Risk Demographics: Approximately 60% of detected security risks originate from non-technical departments. Finance teams rank as the single largest source of vulnerable vibe-coded applications, followed by marketing and legal units, with software developers ranking fourth.
- Speed of Adoption: Newly released commercial and open-source AI tools routinely infiltrate enterprise networks within two hours of public launch, frequently exposing sensitive corporate data to unauthenticated web access.
- Inbound Pipeline: The startup's free component scanner, Plutonium, lets security professionals evaluate public AI connectors and MCP servers, driving 70% of Pluto's inbound enterprise pipeline.
What This Means
The rapid expansion of artificial intelligence in corporate environments has rendered traditional endpoint and network security paradigms insufficient. Legacy security software operates on static rules designed for human user behavior, assuming predictable access patterns and manual software deployment. However, autonomous agents continuously evaluate context, invoke remote tool APIs, and generate dynamic code execution sequences.
When finance or marketing teams use AI platforms to vibe-code internal tools, they often bypass standard software development lifecycle (SDLC) security controls. A single misconfigured API key, permission scope, or public endpoint in a vibe-coded tool can grant threat actors direct access to proprietary customer records, financial reports, or intellectual property. Pluto Security addresses this gap by shifting governance from passive user monitoring to active, real-time agent supervision.
Technical Breakdown
Pluto Security deploys lightweight sensors across endpoint devices, integrating directly with existing endpoint detection and response (EDR) platforms and device management systems. Rather than blocking AI activity outright, the platform monitors agent interactions at runtime to safeguard operational integrity.
- Endpoint Sensor Deployment: Distributes compact monitoring agents directly to workstations to collect running process metadata, active system permissions, terminal commands, and prompt structures.
- Real-Time Context Injection: Intercepts active AI agent sessions attempting unauthorized privilege escalation and injects real-time counter-instructions to force user re-authentication or scope limitation.
- Component and MCP Scanning: Evaluates the security posture of third-party Model Context Protocol (MCP) servers, browser extensions, and external tool skills before execution.
- Local Anomaly Detection: Runs lightweight local models on user devices to evaluate intent and detect malicious prompt injections or exfiltration routines without transmitting sensitive transcripts to external clouds.
Industry Impact
The rise of agentic security startups highlights a broader shift in how enterprises approach artificial intelligence governance. As frontier labs release increasingly capable models, the primary operational bottleneck is no longer raw intelligence, but safe containment and control.
By establishing active runtime intervention capabilities, security teams can permit rapid internal AI experimentation without exposing systems to catastrophic data breaches. Organizations that adopt proactive agent monitoring can safely harness non-technical vibe-coding while mitigating compliance violations, unauthorized cloud spend, and third-party supply chain vulnerabilities.
Looking Ahead
As autonomous swarms and persistent background agents become standard components of enterprise infrastructure, real-time agent governance will become a mandatory requirement for corporate compliance. Security architectures must evolve to treat AI agents as semi-autonomous digital entities requiring granular identity management, behavioral boundaries, and immediate intervention mechanisms.
Pluto Security plans to use its $42 million capital injection to scale its U.S. and Israeli research teams, expand its Plutonium component catalog, and deepen integration with major enterprise AI platforms. CISOs should prepare for a future where securing artificial intelligence requires monitoring not just what humans type into prompt boxes, but what autonomous agents execute behind corporate firewalls.
Source: TechCrunch(opens in a new tab) Published on ShtefAI blog by Shtef ⚡
