Jared Kaplan is a co-founder and the Chief Science Officer of Anthropic, where he holds overall research leadership (Pacing the Frontier (statement from employees of frontier AI companies, July 2026)). A theoretical physicist by training, he is the first author of the 2020 paper "Scaling Laws for Neural Language Models," which established the power-law relationships between model size, dataset size, compute and loss that informed the frontier-AI industry's scaling approach.
Background
Kaplan trained as a theoretical physicist at Johns Hopkins and was on its faculty; the scaling-laws paper carries his Johns Hopkins affiliation alongside OpenAI (Scaling Laws for Neural Language Models). He joined OpenAI in 2019 and co-founded Anthropic in 2021 (OpenAI).
Research
Scaling laws
Kaplan is the first author of "Scaling Laws for Neural Language Models," submitted to arXiv on January 23, 2020 as arXiv:2001.08361, with Sam McCandlish, Tom Henighan, Tom B. Brown, Benjamin Chess, Rewon Child, Scott Gray, Alec Radford, Jeffrey Wu and Dario Amodei; McCandlish is listed as the submitting author (Scaling Laws for Neural Language Models). The paper reports that cross-entropy loss scales as a power law with each of training compute, dataset size and parameter count taken individually, and characterizes those relationships as predictable enough that performance can be forecast before training (Scaling Laws for Neural Language Models). The concept is covered at Scaling Laws.
He is also a co-author of the InstructGPT paper (Training Language Models to Follow Instructions with Human Feedback (InstructGPT)).
Alignment and model specification
Kaplan is among the authors of the Constitutional AI paper, alongside Bai, Kadavath, Kundu, Askell and others with Dario Amodei, Sam McCandlish and Tom Brown (Constitutional AI: Harmlessness from AI Feedback). He is listed as a major contributor to Claude's Constitution, Anthropic's published specification for the values and behavior of its Claude models, which was authored primarily by Amanda Askell with Joe Carlsmith as a significant contributor and Kaplan, Chris Olah and Holden Karnofsky as major contributors (Claude's Constitution).
Role at Anthropic
Kaplan appears as Chief Science Officer in Anthropic's leadership (Anthropic). Observer reported in April 2026 that he has also served as the company's responsible scaling officer since 2024, a role Anthropic's responsible scaling policy defines but does not name a holder for in the published text (Source: observer.com). The RSP assigns that officer, jointly with the chief executive, the task of sharing an approved Risk Report and the internal feedback on it with the Board and the Long-Term Benefit Trust (Anthropic Responsible Scaling Policy v3.4 (July 2026)). The single-source status of the appointment is noted; Anthropic has not published a corresponding confirmation.
He has spoken publicly for the company's technical positions. In a TechCrunch interview published June 6, 2025, "Inside Anthropic's AI ambitions with Jared Kaplan," he set out Anthropic's research direction. At the April 7, 2026 disclosure of Claude Mythos Preview he told The New York Times that the model "demonstrates what is now possible for defenders at scale, and adversaries," framing the release as intended in part "to raise awareness" (Source: nytimes.com). The capability claims from that release are covered at Claude Mythos Preview and AI and Cybersecurity.
Positions
Kaplan was among the first thirty listed signatories of "Pacing the Frontier," the July 28, 2026 statement asking the U.S. government to support an international effort to develop the technical and governance tools needed to deliberately pace frontier AI development. He signed as Co-Founder and Chief Science Officer, one of four Anthropic co-founders to sign in a personal capacity alongside Dario Amodei, Jack Clark and Benjamin Mann, and one of four chief-scientist-equivalent research leads across the frontier developers to appear in that group, with John Schulman of Thinking Machines, Jakub Pachocki of OpenAI and Shengjia Zhao of Meta AI (Pacing the Frontier (statement from employees of frontier AI companies, July 2026)).
Relationships
- co-founded: Anthropic — co-founder and Chief Science Officer.
- related: OpenAI — researcher there from 2019 until co-founding Anthropic.
- supports: Scaling Laws for Neural Language Models, Training Language Models to Follow Instructions with Human Feedback (InstructGPT), Constitutional AI: Harmlessness from AI Feedback, Claude's Constitution — author or contributor.
- supports: Pacing the Frontier (statement from employees of frontier AI companies, July 2026) — signatory.
- depends-on: Scaling Laws — the relationship his best-known paper established.
- related: Dario Amodei, Jack Clark, Chris Olah, Tom Brown — Anthropic co-founders and co-authors.
- related: Johns Hopkins University — former faculty affiliation.