The 1980s were a far cry from what life looks like today. If you weren’t around, you probably don’t remember using paper maps to navigate, renting VHS movies at Blockbuster, printing photos from film, or calling people to talk on their landline phones.
You may also not know that cybersecurity was a thing back then – but even that looked much different than it does today. The early hackers (yes, I was one of them) were motivated by curiosity and the thrill of testing system boundaries. Worries about large-scale cybercrime or widespread threats were largely absent.
Founder & CTO of DataKrypto.
Today, the stakes are dramatically higher. Data has become the crown jewel of every organization and a high-value target of the most ambitious cybercriminals. As such, protecting data is at the heart and mind of every enterprise CISO, and the number one priority of every security team.
Charged with monitoring, defending, and responding to threats, these teams today face unprecedented challenges as the rapid adoption of artificial intelligence (AI) expands the digital attack surface at an accelerated rate.
From siloes to consolidation: AI and the new concentration of data
In the past, enterprise data was spread across multiple systems and silos. While this created operational inefficiencies, it also provided a natural layer of defense, as attackers had to work harder to locate, access, and extract valuable datasets, much like the logistical challenge of moving tons of gold from Fort Knox.
AI has upended this model. Training and inference require large volumes of data to be consolidated and distilled into compact, portable models. Instead of stitching together fragments from dispersed systems, attackers now face concentrated repositories of sensitive information that are far easier to target and penetrate.
With little built-in security, AI leaves data exposed “in the clear,” making the defender’s job less like guarding Fort Knox and more like protecting a pouch of diamonds – smaller, denser, and far easier to steal.
This scenario illustrates why companies using AI face a significantly greater risk of data breaches, leading to financial and reputational harm and compliance failures.
In fact, research suggests that more than three-quarters of businesses have already experienced AI-related breaches, ranging from accidental data leakage to deliberate poisoning of training datasets, which jeopardize their compliance efforts with regulatory frameworks like the General Data Protection Regulation (GDPR), the Health Insurance Portability and Accountability Act (HIPAA), and the California Consumer Privacy Act (CCPA).
Without proper safeguards, the risks of using AI are clear, and may be contributing to problems identified in the recent MIT report stating that enterprises are not getting full value from their AI investments.
When the stakes are so high, companies are limiting their use of AI, which also limits their ability to create innovative solutions to highly complex challenges.
When it comes to AI, traditional data security methods are no longer enough.
A new approach: Continuous encryption
Conventional encryption focuses on protecting data at rest (when stored) and in transit (when moved across networks). But in the age of AI, the most dangerous gap lies with data in use — the moment it is loaded into memory for processing, when it must be decrypted or made “in clear.”
This is where continuous encryption becomes indispensable. By maintaining encryption across the entire lifecycle, from storage and transmission through active computation, sensitive information remains shielded at all times. Even during training or inference, data is never decrypted or made vulnerable to unauthorized access.
Two crucial technologies make this possible:
- Fully Homomorphic Encryption (FHE): Allows computations directly on encrypted data, ensuring that raw values never need to be decrypted.
- Confidential Computing with Trusted Execution Environments (TEEs): Provides secure enclaves where temporary operations occur in isolated, protected memory, inaccessible even to system administrators or cloud providers.
Combined, these technologies create a “zero-knowledge” environment in which neither the AI provider nor malicious actors can reconstruct inputs, outputs, or models outside the secure enclave. As a result, both customized open-source and proprietary AI LLMs are fully protected, while ensuring the privacy and security of sensitive data.
This zero-knowledge AI environment prevents a range of significant data exposure risks, including the unintentional leakage of sensitive information by employees using both sanctioned and unsanctioned generative AI tools.
A new mandate: End-to-end protection across use cases
For security teams, continuous encryption offers a practical way to safeguard both sensitive data and the models trained on it. This approach presents a new mandate: protection must extend beyond storage and transmission to encompass every stage of data use in the AI lifecycle. The benefits across industries and use cases are powerful:
- Healthcare: Patient records can be analyzed for predictive insights without risking exposure of personal health information.
- Financial services: Fraud detection and risk assessment models can run on encrypted customer data without jeopardizing privacy or compliance.
- Public sector and critical infrastructure: Agencies can share intelligence securely, knowing that sensitive information stays protected throughout the analytics process
- Retail & consumer services: Retailers can leverage AI to personalize shopping experiences and loyalty programs while protecting customer purchase histories and personal data.
- Telecommunications & cloud: Providers can optimize networks and deliver secure multi-tenant AI services without risking customer data exposure.
In each case, encryption ensures that sensitive data remains inaccessible to unauthorized parties, even if models are stolen or environments are compromised.
Aligning security with innovation
The rise of AI doesn’t have to come at the expense of security or compliance. Continuous encryption allows organizations to harness the power of AI while maintaining confidentiality, integrity, and regulatory alignment.
This method of end-to-end data protection means organizations can close the most critical security gap in AI, giving them the confidence to innovate without fear of exposure.
We’ve featured the best endpoint protection software.
This article was produced as part of TechRadarPro’s Expert Insights channel where we feature the best and brightest minds in the technology industry today. The views expressed here are those of the author and are not necessarily those of TechRadarPro or Future plc. If you are interested in contributing find out more here: https://www.techradar.com/news/submit-your-story-to-techradar-pro
https://cdn.mos.cms.futurecdn.net/fg7bgy65pWhFo4Qzib58yX-2560-80.jpg
Source link
