“In terms of buyer knowledge, we provide a flat assure to our prospects: buyer knowledge is buyer knowledge. We by no means use that knowledge for coaching any mannequin ourselves. All of our prospects can relaxation assured that their knowledge, and any AI merchandise they construct on prime of it, will solely be used to reply questions for them,” Snowflake CEO Sridhar Ramaswamy informed indianexpress.com.
On the newest version of Snowflake Summit, Ramaswamy highlighted the corporate’s dependable dedication to knowledge safety and accountable AI implementation. The CEO spoke concerning the complete measures the corporate is endeavor to guard buyer knowledge whereas enabling progressive AI capabilities.
Addressing the considerations about AI corporations and their knowledge utilization practices, Ramaswamy mentioned, “We’ve constructed mechanisms that floor safety knowledge and alerts to our prospects, together with visibility into what number of of their accounts aren’t protected with sturdy authentication. We take these classes severely and are devoted to actively collaborating with our prospects to make sure the safety of their knowledge.
The AI knowledge cloud firm has, over time, strengthened its safety posture, implementing what it describes as a ‘shared future’ mannequin that locations better duty on Snowflake to proactively defend prospects moderately than relying solely on conventional shared duty frameworks. Integral to Snowflake’s enhanced safety technique is the obligatory implementation of two-factor authentication throughout all accounts.
“We introduced that we wish to make it obligatory for each account to have two-factor authentication,” Ramaswamy defined, including that the corporate has additionally built-in fashionable authentication strategies resembling Face ID and different biometric applied sciences.
Brad Jones, Snowflake’s chief data safety officer and vice chairman of data safety, elaborated on the corporate’s aggressive stance on eliminating single-factor authentication fully. “We finally consider to be essentially the most safe, now we have to remove a few of these legacy authentication strategies from our prospects to actually get to that finest posture for them.”
The safety enhancements prolong past primary authentication. Snowflake has launched passwordless authentication choices, together with passkeys, programmatic entry tokens, and assist for numerous authenticators like Google Authenticator, transferring away from its conventional reliance on Duo for multi-factor authentication.
Story continues under this advert
Jones shared that Snowflake has applied the ‘leak password safety’ – a system that actively displays the darkish internet for compromised credentials related to Snowflake accounts. “We’re wanting at nighttime internet for credentials, usernames and passwords which have been leaked on the market by way of numerous means like info-stealing malware or different strategies, and we’re validating if these are lively credentials which are lively in a Snowflake account. We’ll go so far as locking that account and actually asking questions later,” Jones defined.
AI and safety
Through the years, AI has change into integral to Snowflake’s platform choices. Contemplating the meteoric tempo of developments in AI, Snowflake has established complete governance frameworks to make sure accountable deployment. Jones knowledgeable that the corporate not too long ago achieved ISO 42001 certification, an AI governance commonplace.
“We not too long ago bought the ISO 42001 certification, which is the AI governance certification based mostly on the ISO commonplace that got here out. I consider we’re one of many first cloud corporations, if not the primary type of cloud firm, to get that,” Jones famous, explaining that this required growing commonplace insurance policies and processes for acceptable AI use.
Snowflake’s strategy to AI safety makes use of its unified platform structure, guaranteeing that current knowledge governance and safety controls robotically prolong to AI workloads. “As a result of we’re a single, unified platform, all the guidelines that you’ve got, whether or not it’s for knowledge entry or knowledge masking, work out of the field with the AI merchandise that you simply create on prime,” Ramaswamy defined.
Story continues under this advert
“In fact, AI brings challenges like immediate injection and adversarial enter. To counter that, we provide options like Cortex Guardrails to guard towards malicious prompts and guarantee accountable behaviour,” the CEO added.
Addressing evolving risk panorama
Snowflake acknowledges that for cybersecurity AI presents each alternatives and challenges. “AI, everybody sees it as a double-edged sword, proper? Any firm that doesn’t lean into it and undertake it’s going to be left behind, however it’s important to put sturdy guardrails and tips in place,” Jones famous.
To handle AI-enhanced threats, Snowflake employs complete purple group testing and maintains sturdy detection capabilities. “I don’t consider there are any new, novel assaults which are happening on account of AI. It’s the identical issues, however higher and sooner. In order that’s incumbent then on safety groups to have the ability to react faster,” Jones mentioned.
The corporate continues to put money into superior safety capabilities, together with machine studying fashions for detecting suspicious behaviour and enhanced visibility instruments by way of its Belief Centre platform. “We’re by no means going to cease on this space; as we discover issues which are going to be helpful for our prospects that we will use internally to assist them or present on to them, we are going to proceed to innovate,” Jones concluded.