Get Plus

Sign in to baba News

One account across the web, iPhone and Android — your subscription follows it.

or use an email code

Search stories

Type at least two characters. Results come from every newsroom baba reads.

↑↓ to move · ↵ to open · esc to close

Sign in to baba News

Sign in to keep asking. News Plus removes the daily limit.

or use an email code

Ongoing Story· Day 2

Israeli AI Startup Plurai Raises $13 Million for Automated Agent Testing

2 developments

NEWSru IsraelTech

Israeli Startup Plurai Develops AI Agent Testing Platform

Translated & summarized from NEWSru Israel by baba

CommunityNeutral tone

Russian · 1 newsroom

Israeli startup Plurai has launched a platform to test AI agents by simulating thousands of virtual customers trying to break them. The system uses a 'rule map' to generate complex test scenarios, inspired by autonomous vehicle testing. Plurai also offers an AI-trained 'little guardian' for real-time monitoring, claiming it's more accurate and cheaper than commercial models. The company, founded in 2025, has 12 employees and secured $13 million in seed funding.

The story in 6 lines · by baba

  • Plurai's new platform tests AI agents using thousands of simulated customers trying to exploit vulnerabilities.
  • The system generates complex test scenarios based on a 'rule map' of AI agent regulations.
  • Plurai's approach is inspired by simulation methods used in the development of autonomous vehicles.
  • The company also developed an AI-trained 'little guardian' for real-time monitoring of AI agents.
  • Plurai, founded in 2025, has 12 employees and raised $13 million in seed funding.
  • The AI agent testing platform aims to prevent costly errors and unexpected failures in AI systems.

Israeli startup Plurai has developed a platform designed to rigorously test AI agents by simulating thousands of virtual customers attempting to exploit vulnerabilities and disrupt their operations. These AI agents are sophisticated programs that go beyond simple question-answering, engaging in tasks like processing returns, modifying bookings, and managing customer data. Each agent must adhere to numerous rules, and any errors can be costly. The challenge lies in the unpredictable interactions between these rules, making manual testing of all possible scenarios infeasible. Plurai's co-founder and CEO, Ilan Kadar, noted that AI systems often fail when encountering unforeseen difficulties. The company's approach draws inspiration from the autonomous vehicle industry, where Kadar and his partner, mathematician Elad Levy, previously worked at Nexar on systems for Waymo. Kadar highlighted that Waymo achieved customer trust by accumulating extensive simulation mileage, far exceeding real-world driving. Plurai applies a similar methodology to AI agents, creating a digital twin of an agent with its knowledge base, tools, and company rules. This digital twin is then subjected to tests by thousands of virtual customers, including those designed to deceive the agent and bypass its limitations. The platform's testing scenarios are generated using a 'rule map' that identifies commonly co-occurring rules. This allows the system to select combinations of increasing complexity without exhaustively testing every single possibility. Plurai claims this method yields test results comparable to human-curated benchmark datasets. Additionally, Plurai proposes a 'little guardian' system for real-time agent monitoring. Training this guardian requires thousands of labeled examples, which Plurai generates without human intervention by having one AI model create complex scenarios and others evaluate their suitability for testing. According to the company, this AI-trained guardian performs more accurately and cost-effectively than larger commercial models. Founded in 2025, Plurai has 12 employees in Israel and the US and has raised $13 million in seed funding.

NEWSru IsraelOther · Tel Aviv

Sign in to baba News

Sign in to follow newsrooms, topics and people.

or use an email code

TechTopic

Sign in to baba News

Sign in to follow newsrooms, topics and people.

or use an email code

Mentioned

Related stories · 1

Not the same event — other stories that share this one’s people, places, or theme: background, reactions, and follow-ups.

Open the Wire