Innovation That Matters

| Photo source © Kevin Ku from Pexels via Canva.com

Fighting cyber threats in the age of gen AI

Computing & Tech

How can this tool help companies protect themselves against evolving AI attacks?

Spotted: According to research firm Gartner, by 2026, more than 80 per cent of businesses will have used generative AI in some capacity. This technology may have the potential to revolutionise the way companies across every sector are run, but it’s also transforming the capabilities of hackers and poses an increasing cybersecurity risk. Now, Lakera is offering businesses AI-powered cyber-protection.

Traditional cybersecurity processes typically involve strict and static rules that fail to respond to the adaptability of generative AI, but by employing AI itself, Lakera helps to stay ahead of cyber threats and prevent data leaks, without slowing down or compromising the experience for the end consumer.

The startup’s low-latency AI application firewall, called Lakera Guard, can be installed in a few minutes with just one line of code and detects and blocks cyber attacks against a business’ generative AI applications – including customer-facing chatbots – in real time. This means protecting against prompt injection, where hackers attempt to trick large language models (LLMs) using prompts to elicit a harmful response – including receiving unauthorised access. Lakera Guard learns to differentiate normal and harmful prompts based on the specific use case.

As well as blocking harmful prompts, Lakera Guard also acts as an automated AI moderator, ensuring that users aren’t shown inappropriate content. As well as protecting customers from malicious content and misinformation, this also helps businesses stay compliant and avoid potential reputational damage.

Beyond its comprehensive security platform, Lakera has also created Gandalf, a language model educational game that gives users better insight into the potential vulnerabilities of AI applications, by playing the role of the hacker who must guess Gandalf’s passwords. Lakera uses insights from Gandalf, which has over one million players and 35 million attack data points (and growing), as well as public data and internal information, to teach its cybersecurity model so that it responds effectively to the latest AI threats.

In July this year, the company raised €18.4 million in a series A financing round led by venture capital firm Atomico. With the latest, sizable funding, Lakera plans to develop its product offering further as well as boost its presence across more of the US.

Written By: Matilda Cox

Website: lakera.ai

Contact: lakera.ai/contact