Insights on Data Governance, AI Governance, AI Products, Data Architecture, and Product Strategy. By Vikas Pratap Singh.
Part 6 showed Meridian adopting differential privacy for query analytics. This article explains why: what epsilon means, how noise is calibrated, what Apple and the Census Bureau chose, and when to use synthetic data instead.
Part 6 showed Meridian rejecting federated learning (single-tenant architecture) and deferring homomorphic encryption (47x latency). This article explains the mechanics behind those decisions, introduces secure multi-party computation, and reveals the tension between GDPR's explainability mandate and privacy protection. Concludes with a capstone PET decision framework spanning Parts 7 through 9.
Part 3 introduced PETs as governance decisions. Part 6 showed Meridian evaluating them. This article explains how each technique actually works: static and dynamic masking, vault-based and format-preserving tokenization, and the k-anonymity family of de-identification methods.
A study of 32 models across 56 jailbreak techniques found attack success rates jumping from 0.6% to 96.3% depending on the attack type. The Safety Tax costs 7-31% accuracy. Ten alignment techniques are competing to solve this. None covers the full attack surface.
Fine-tuning GPT-3.5 on 10 examples for $0.20 strips its safety guardrails. Removing safety from Llama 3 takes 5 minutes on one GPU. This article explains the mechanism: safety alignment concentrates in the first few output tokens, creating a shallow defense that fine-tuning, prefilling attacks, and adversarial suffixes bypass trivially.
The EU AI Act makes you responsible for safety when you fine-tune. Reasoning models can autonomously jailbreak other models at 97% success. Half of organizations have no formal AI guardrails. This article provides the regulatory map, the liability analysis, and a minimum viable safety governance checklist.
Essays on thinking clearly, living intentionally, and building a career worth having.
Current AI can assist with routine cognitive work, but accountable judgment, creative synthesis, and sustained concentration remain human strengths. Deep work is now the scarcest, most valuable skill a knowledge worker can build.
78% of workers say meetings prevent real work. Camera-on can increase fatigue, while camera-off can reduce how engaged you appear. The problem is not knowledge. It is that obvious best practices feel optional under pressure.
In a world of seven-hour daily screen averages and optimization culture, the hobbies that feel unproductive are the ones your brain needs most. The neuroscience explains why.
Get new articles on data governance, AI, and engineering delivered to your inbox.
No spam. Unsubscribe anytime.