Miju Labs

← All buyers · ElevenLabs

Artemis Seaford

Vice President of AI Safety, ElevenLabs · ElevenLabs

medium confidenceTier B · Map and watchImage, video & audio modelsaudio

Leads AI safety at ElevenLabs — a team of roughly seventeen — covering misuse prevention for voice and audio models, including consent and verification for voice data. She is a gatekeeper rather than a buyer: any human-data programme touching voices, likeness or performance has to satisfy her function.

Background

An unusually credentialled policy operator: PhD in Political Science and a JD from Stanford, an MPP from the Harvard Kennedy School and a first-class BA from Oxford, with earlier work in migration policy and refugee response. She ran usage policies for ChatGPT and its API at OpenAI and worked at Meta on countering harassment, espionage and threats to user safety before joining ElevenLabs.

She is a visible public speaker on AI ethics — TechCrunch Sessions: AI in May 2025 alongside Ion Stoica, the ITU AI for Good programme, and an announced speaker at the AE Global Summit on Open Problems for AI in 2026. Her public subject is synthetic-media misuse and detection rather than model quality.

What they run now

Preventing and mitigating misuse of ElevenLabs' audio and voice models: consent and verification for voice data, detection of synthetic audio (the company ships an AI Speech Classifier), and policy for a product line whose core risk is impersonation.

What they buy

Inference: red-teaming and misuse-probe datasets, consent-verification processes for contributed voices, and safety evaluation of model outputs. Not aesthetic judgement. Her function is a constraint on data acquisition — what can be collected, from whom, with what consent — more than a source of demand for it.

Career

  1. presentVP of AI Safety, ElevenLabs — ~17 people per The Org; also described as AI Safety Team Lead
  2. —Usage policies for ChatGPT and the API, OpenAI
  3. —Countering harassment, espionage and threats to user safety, Meta
  4. —Policy work before tech, Migration policy / refugee response
  5. —Stanford University, PhD, Political Science
  6. —Stanford University, JD
  7. —Harvard Kennedy School, Master of Public Policy
  8. —University of Oxford, BA (First Class)

On the record

AI ethics and misuse · 2025-05

Spoke on where AI crosses the line and the ethical crisis facing deployed AI systems, alongside Ion Stoica. source

Interviews & talks

Why it matters here

Little, directly. She does not buy creative-taste data and ElevenLabs is audio. She matters in two indirect ways: she is the person who would veto or shape a voice/likeness data deal, so any ElevenLabs engagement has to be consent-clean from the start; and her consent-and-provenance framing is the same argument Adobe uses about Stock contributors, so she is a useful public reference point when Julian needs to show that paid, consented human contribution is the industry norm.

How to reach

Speculation: not a sales target. If contact happens it should be at a conference — she speaks frequently — and on the topic of consented, compensated human contribution to training data, where a creative-data vendor and a safety lead have genuinely aligned interests. She is active on X.

What we could not establish
  • Her exact title varies across sources: VP of AI Safety, AI Safety Team Lead, and 'VP of AI Security' on the ITU page.
  • Whether her remit covers data-acquisition consent formally or only misuse.
  • Exact dates for the OpenAI and Meta roles.

Why they are on the map

Runs AI safety (~17 people) — consent/verification of voice data. source

Runs vendor projects; decides if a pilot converts.

How ElevenLabs buys data

Public 'Voice Data Partnerships' program buys conversations, audiobooks, character performances and news audio for base-model training or fine-tuning. Data Operations job: owns end-to-end data acquisition and labelling, 'manage external data vendors', oversee in-house and third-party labelling teams. Voice Curator role (removed Jan 2026) had staff judge voice naturalness/expressiveness. Research Engineer postings cite post-training, data curation and benchmark design.

Research & Research Engineering under co-founder/CTO Piotr Dąbkowski; a Data Ops team sits beside Research translating needs into labelling specs and vendor work; AI Safety org (~17) under VP Artemis Seaford.

Known vendors: unnamed external labelling vendors (per Data Ops job), voice data partners (program)

Sources

  1. ITU AI for Good speaker profile (career and education)
  2. The Org: Artemis Seaford, VP of AI Safety at ElevenLabs
  3. TechCrunch Sessions: AI (May 2025)
  4. X post announcing her as AE Global Summit speaker, naming her VP of AI Safety at ElevenLabs