SHELFCRITTER AI Digest

COMPANY

Sampura Research

companytopic-noteai-safetynonprofit

Overview

Sampura Research is a London-based AI-safety nonprofit co-founded by ex-Google DeepMind researcher Rishub Jain with Joshua Jacob and Alex Adams, launched publicly in August 2026 with a $6.5M raise plus $4.2M pledged. Founding pitch is architecture of human-in-the-loop primitives at the model layer — a “human-judge hybrid” scalable-oversight system — positioned against what the founders characterise as the “rogue agents” framing pushed by frontier-lab CEOs. Sits inside the already-crowded scalable-oversight problem space alongside METR, Redwood, Apollo, and Anthropic’s own alignment team.

Timeline

  • 2026-08-26-AI-Digest — Sampura Research launches as a live, funded AI-safety nonprofit (Bloomberg / EdTech Innovation Hub) — London-based, co-founded by ex-DeepMind researcher Rishub Jain with Joshua Jacob and Alex Adams; $6.5M raise plus $4.2M pledged. Founding technical pitch: architected human-in-the-loop primitives at the model layer (a “human-judge hybrid” scalable-oversight system), positioned against the “rogue agents” framing pushed by frontier-lab CEOs. Narrow read the digest carries: this is a live, funded launch — not an announcement of intent, with a specific technical thesis (model-layer primitives, not policy-layer guardrails). Load-bearing framing the digest carries: do NOT frame this as AI safety “splintering into a third pole” or as an anti-lab counterweight — the scalable-oversight problem space is already crowded (METR, Redwood, Apollo, Anthropic alignment team all overlap with Sampura’s pitch to varying degrees), and $6.5M is small next to those established players’ operating budgets. Right frame is accretion — another specific technical bet on human-in-loop oversight, not a new pole.

Key Developments

  1. Live Funded Launch With Human-Judge-Hybrid Model-Layer Primitives Pitch (August 26, 2026): Sampura’s launch is $6.5M raised plus $4.2M pledged behind a specific technical thesis — architected human-in-the-loop primitives at the model layer, framed by the founders as a “human-judge hybrid” scalable-oversight system positioned against the “rogue agents” framing pushed by frontier-lab CEOs. Founding team: ex-DeepMind Rishub Jain, Joshua Jacob, Alex Adams. Load-bearing framing to carry: live, funded launch — not an announcement of intent; the technical thesis (model-layer primitives, not policy-layer guardrails) is what distinguishes it from adjacent scalable-oversight programs (METR, Redwood, Apollo, Anthropic alignment team). Structural read: accretion in an already-crowded space, not a new safety pole — $6.5M sits below the operating budgets of established players; the interesting question is whether the human-judge-hybrid primitive gets picked up by any frontier lab as an integration point rather than a competing framing. 30 / 60 / 90-day watch: whether Sampura publishes its human-judge-hybrid architecture specification; whether any frontier lab (Anthropic, OpenAI, DeepMind) integrates or references the primitive; whether a second London-based AI-safety nonprofit launches inside the same ex-DeepMind talent-flow window.

See also: Anthropic, DeepMind, MOC - Agent Security, MOC - Major Companies.