Bedrock Security Takes On Data Complexities in GenAI Era

In the rapidly evolving world of cloud computing and generative artificial intelligence (AI), data security has become a paramount concern for organizations worldwide. Bedrock Security, a newly launched platform, aims to address these challenges head-on.

Bedrock Security has secured a $10 million seed funding from leading cybersecurity investor Greylock, underscoring the vital importance of data in driving organizational growth. The company’s founders are acutely aware of the need to protect this data against the risks posed by emerging technologies.

“We purpose-built Bedrock with the imperative understanding that data is foundational to organizational growth, especially with the rapid adoption of generative AI,” said Pranava Adduri, CEO and co-founder of Bedrock Security, in a Tuesday (March 26) news release announcing the debut of Bedrock’s data security platform.

The Challenges of AI Data

Bedrock provides protection for sensitive data in industries like banking, insurance, healthcare, retail and financial services, where keeping regulated data safe is crucial. Additionally, Bedrock caters to specialized areas such as genomics, electronic design automation (EDA) and computer-aided design (CAD) because it can handle unique, proprietary formats of intellectual property.

The company’s solution centers around its AI Reasoning (AIR) Engine. AIR goes beyond the limitations of traditional rule-based systems, and is continuously learning and adapting to understand the true context and sensitivity of a company’s data.

Adduri said in an interview with PYMNTS that this approach is essential for industries operating under strict regulations and those handling valuable intellectual property (IP). 

“Traditional approaches to data security rely on rules for identifying regulated data and intellectual property,” he said. “However, rules don’t work when your data changes and new data types are introduced, especially in today’s explosive data growth for cloud and GenAI. AI Reasoning (AIR) uses the latest in AI to reason about what data is most material and important to a business and allows Bedrock to keep pace with new data as it’s being created.” 

AI is revolutionizing how security teams tackle cyber threats, enhancing their speed and efficiency. It sifts through massive datasets to spot complex patterns, automating the preliminary steps of incident analysis. This advancement provides security experts with a comprehensive view, significantly reducing the time it takes to respond.

Timothy E. Bates, Lenovo’s former CTO, highlighted AI’s role in bolstering defenses.

“Machine learning helps in spotting unusual activities, and AI-powered platforms deliver in-depth threat insights and predictive analysis,” he said. 

Bates also pointed out the effectiveness of deep learning in dissecting malware to decode its structure and possibly reverse-engineer attacks.

“These AI solutions operate behind the scenes, constantly evolving with every new threat, aiming not only to protect but also to neutralize future dangers,” he added. 

As the global economy becomes increasingly interconnected, cybercrime is on the rise. According to an FBI report, the U.S. alone faced over $10.3 billion in losses due to cyberattacks in 2022.

Keeping IP Safe

To keep sensitive data and core intellectual property (IP) safe from accidentally being used in large language models (LLMs), it’s essential first to know exactly what and where this data is, Adduri said. That’s where Bedrock Security’s AIR comes into play.  

Bedrock also uses Trust Boundary data perimeters, which act like safety zones that define where this sensitive information can be used. These Trust Boundaries help businesses monitor their important data and IP, making sure it doesn’t unintentionally end up in LLM training sessions or get pulled into systems that use a lot of data to answer questions. 

Dealing with data risks means staying updated with every piece of data a company makes, including how it’s used and shared, Adduri noted. Traditional methods struggle to keep up with the vast amounts of data we have today.

“Since Adaptive Sampling allows for continuous discovery and Trust Boundaries allow you to specify how your regulated data and core IP may be used and accessed, any violations of controls are detected near real-time, allowing security teams to rapidly spring to containment versus waiting for leaks to become breaches or fail compliance control audits,” he said.