Highlights:

  • The flagship product of Laiyer AI, LLM Guard, is tailored to offer security solutions for large language models like OpenAI LP’s GPT-4, addressing various security threats, misuse, and potential attacks.
  • As per Laiyer, LLM Guard delivers a significant threefold reduction in CPU inference latency. This advancement allows for utilizing cost-effective CPU instances instead of pricier GPUs, all while maintaining accuracy levels.

Recently, Protect AI Inc., a cybersecurity firm specializing in artificial intelligence and machine learning, revealed its acquisition of the artificial intelligence security startup Laiyer AI for an undisclosed amount.

The flagship product of Laiyer AI, LLM Guard, is tailored to offer security solutions for large language models like OpenAI LP’s GPT-4, addressing various security threats, misuse, and potential attacks. LLM Guard shields against several threats, including prompt injection attacks, data poisoning, and multiple vulnerabilities that may emerge within AI models.

Prompt injection attacks entail attackers manipulating LLMs through crafted inputs, potentially resulting in data exposure or manipulation of decisions. These attacks can occur by directly manipulating the LLM’s input or indirectly using tainted data sources. They often evade detection due to the implicit trust placed in LLM outputs.

LLM Guard is capable of identifying, sanitizing, and redacting LLM inputs and outputs. By doing so, the potential for data breaches caused by manipulated model outputs is reduced, thereby safeguarding the confidentiality and integrity of the data that these models process. The service is designed to enhance performance and seamlessly integrate with pre-existing security procedures. It provides observability tools, including logging and metrics.

As per Laiyer, LLM Guard delivers a significant threefold reduction in CPU inference latency. This advancement allows for utilizing cost-effective CPU instances instead of pricier GPUs, all while maintaining accuracy levels. The code is open-source and has garnered considerable adoption, with Laiyer AI witnessing over 13,000 library downloads and 2.5 million downloads of its models on Hugging Face within 30 days.

Following the acquisition, Protect AI aims to develop a commercial version of LLM Guard enriched with expanded features, capabilities, and integrations seamlessly integrated within the Protect AI platform. “These new capabilities will empower our customers in automotive, energy, manufacturing, life sciences, financial services, and government sectors to develop safe, secure GenAI applications,” Ian Swanson, the Chief Executive, stated before the disclosure.

Neal Swaelens and Oleksandr Yaremchuk, the Co-founders of Laiyer, stated that “By joining forces with Protect AI, we are extending Protect AI’s products with LLM security capabilities to deliver the industry’s most comprehensive end-to-end AI security platform.”

As a venture capital-backed startup, Protect AI secured its latest funding round of USD 35 million in July. Investors in the company comprise Evolution Equity Partners LLP, Salesforce Ventures LLC, Acrew Capital LP, Boldstart Ventures LLC, Knollwood Capital LLC, and Pelion Ventures Partners LLC.