With the surge in both supply & demand for data, individuals are leaving behind increasingly extensive digital footprints, making personal information more vulnerable to misuse or unauthorized access. We have seen cases where personal data gets leaked with scandals like Cambridge Analytica.

For those who are not caught up to speed, check out part 1 of the series where we’ve discussed:

• The importance of data
• Growing demand for data for AI
• The emergence of data layers

Regulations like the GDPR in Europe, California’s CCPA, and others worldwide have made data privacy not just an ethical issue but a legal requirement, pushing companies to ensure data protection.

Given the surge in AI developments, AI plays a pivotal role in both enhancing and further complicating the landscape of privacy & verifiability. For instance, while AI can help detect fraudulent activities, it also enables the creation of deepfakes, making it harder to verify the authenticity of digital content.

The Good

• Privacy-preserving ML: Federated learning allows AI models to be trained directly on devices without centralizing sensitive data, thus preserving user privacy.
• AI can be used to anonymize or pseudonymize data, making it harder to trace back to individuals while still useful for analysis.
• AI is crucial in developing tools to detect and mitigate the spread of deepfakes, ensuring the verifiability of digital content (as well as detecting/verifying the authenticity of AI agents).
• AI can help in automatically ensuring that data handling practices comply with legal standards, making the process of verification more scalable.

The Challenges

• AI systems often require vast datasets to function effectively, but how this data is used, stored, and who has access to it can be opaque, raising privacy concerns.
• With enough data and sophisticated AI, it’s possible to re-identify individuals from supposedly anonymized datasets, undermining privacy efforts.
• With AI capable of generating highly realistic text, images, or videos, distinguishing between authentic and AI-fabricated content becomes harder, challenging verifiability.
• AI models can be tricked or manipulated (adversarial attacks), compromising the verifiability of data or the integrity of AI systems themselves (as seen from Freysa, Jailbreak, etc.).

The challenges have spurred a surge in developments in AI x Blockchain x Verifiability x Privacy, utilizing the strengths of each technology. We’re seeing the rise of:

• Zero-Knowledge Proofs (ZKPs)
• Zero-Knowledge Transport Layer Security (zkTLS)
• Trusted Execution Environment (TEE)
• Fully Homomorphic Encryption (FHE)

1. ZKPs

ZKPs allow one party to prove to another that they know something or that a statement is true without revealing any information beyond the proof itself. AI can leverage this to demonstrate that data processing or decisions meet certain criteria without disclosing the data itself.

A good case study is@getgrass_io""> @getgrass_io. Grass leverages unused internet bandwidth to collect and organize public web data for training AI models.

Grass Network allows users to contribute their idle internet bandwidth through a browser extension or app. This bandwidth is used to scrape public web data, which is then processed into structured datasets suitable for AI training. The network uses nodes run by users to perform this web scraping.

Grass Network emphasizes user privacy by only scraping public data, not personal information. It uses ZKPs to verify and secure the data’s integrity and origin, preventing data corruption and ensuring transparency. This is managed through a sovereign data rollup on the Solana blockchain, which handles all transactions from data collection to processing.

Another good case study is@zkme_""> @zkme_

zkMe’s zkKYC solution addresses the challenge of conducting KYC processes in a privacy-preserving manner. By utilizing ZKPs, zkKYC enables platforms to verify user identities without exposing sensitive personal information, thereby maintaining compliance while safeguarding user privacy.

2. zkTLS

TLS = Standard security protocol that provides privacy and data integrity between two communicating applications (most commonly associated with the “s” in HTTPS).

zk + TLS = Enhancing privacy and security in data transmission.

A good case study is@OpacityNetwork""> @OpacityNetwork

Opacity employs zkTLS to offer secure and private data storage solutions. By integrating zkTLS, Opacity ensures that data transmission between users and storage servers remains confidential and tamper-proof, addressing privacy concerns inherent in traditional cloud storage services.

Use case — Earned Wage Access

Earnifi, an app that has reportedly climbed to a top position in app store rankings, particularly in finance categories, leverages@OpacityNetwork""> @OpacityNetwork‘s zkTLS.

Privacy: Users can prove their income or employment status to lenders or other services without revealing sensitive bank details or personal information like bank statements.

Security: The use of zkTLS ensures that these transactions are secure, verified, and private. It prevents the need for users to trust third parties with their full financial data.

Efficiency: This system reduces the cost and complexity associated with traditional earned wage access platforms that might require extensive verification processes or data sharing.

3. TEE

TEEs provide a hardware-enforced separation between the normal execution environment and a secure one.

Possibly the most well-known security implementation on AI Agents in order to ensure that they’re fully autonomous agents.

Popularized by:

• @123skely"">@123skely‘s@aipool_tee""> @aipool_tee experiment: A TEE pre-sale where a community sends funds to an agent, which autonomously issues tokens based on predefined rules.
• @marvin_tong"">@marvin_tong‘s@PhalaNetwork""> @PhalaNetwork: MEV protection, integration with@ai16zdao""> @ai16zdao‘s ElizaOS, and Agent Kira as a verifiable autonomous AI agent.
• @fleek"">@fleek‘s one-click TEE deployment: Focusing on ease-of-use and accessibility for developers.

4. FHE

A form of encryption that allows computations to be performed directly on encrypted data without needing to decrypt it first.

A good case study is@mindnetwork_xyz""> @mindnetwork_xyz and their proprietary FHE tech/use cases.

Use Case — FHE Restaking Layer & Risk-free Voting

FHE Restaking Layer

By using FHE, restaked assets remain encrypted, meaning private keys are never exposed, significantly reducing security risks. This ensures privacy while verifying transactions.

Risk-Free Voting (MindV)

Governance voting occurs over encrypted data, ensuring votes remain private and secure, reducing coercion or bribery risks. Users earn voting power ($vFHE) by holding restaked assets, decoupling governance from direct asset exposure.

FHE + TEE

By combining TEE and FHE, they create a robust security layer for AI processing:

• TEE shields operations within the computing environment from external threats.
• FHE ensures operations occur on encrypted data throughout the process.

For institutions handling $100mn - $1BN+ in transactions, privacy and security are paramount to prevent frontrunning, hacking, or exposure of trading strategies.

For AI Agents, this double encryption enhances privacy & security, making it useful for:

• Sensitive training data privacy
• Protecting internal model weights (preventing reverse engineering/IP theft)
• User data protection

The main challenge for FHE remains its high overhead cost due to computational intensity, leading to increased energy consumption and latency.

Ongoing research is exploring optimizations such as hardware acceleration, hybrid encryption techniques, and algorithmic improvements to reduce computational burdens and enhance efficiency. Thus, the best use cases for FHE are low computation, high latency applications.

Wrapping Up for Part 2

FHE = Operations on encrypted data w/o decryption (strongest privacy but most expensive)

TEE = Hardware, secure execution in an isolated environment (balance between security & performance)

ZKP = Proving statements or authenticating identities without revealing underlying data (good for proving facts/credentials)

This is a vast topic to cover, so this is not the end. One key question remains: how can we ensure that AI-driven verifiability mechanisms are truly trustworthy in an era of increasing deepfake sophistication? In Part 3, we dive deeper into:

• The verifiability layer
• The role of AI in verifying data integrity
• Future developments in privacy & security

Stay tuned!

Additional Quality Resources on TEE & ZKPs (below)

Disclaimer:

1. This article is reprinted from [0xJeff]. All copyrights belong to the original author [0xJeff]. If there are objections to this reprint, please contact the Gate Learn team, and they will handle it promptly.
2. Liability Disclaimer: The views and opinions expressed in this article are solely those of the author and do not constitute any investment advice.
3. The Gate Learn team does translations of the article into other languages. Copying, distributing, or plagiarizing the translated articles is prohibited unless mentioned.