LASR Labs

What LASR Labs does

LASR Labs (London AI Safety Research Labs) is a 13-week, cohort-based technical AI safety research training programme operated under Arcadia Impact, a registered charity in England and Wales. The programme recruits ML engineers and scientists through a three-stage application process (form, work test, interview) and trains them to conduct publishable AI safety research on loss-of-control risks from advanced AI systems. Participants work in teams of three to four, supervised by experienced AI safety researchers from organizations including Google DeepMind, UK AISI, and top UK universities, producing academic-style papers and accompanying blog posts. The programme runs twice per year in person from the London Initiative for Safe AI (LISA) offices, with plans to expand to three cohorts serving 100+ participants annually, and has recently launched a year-long research programme with UK AISI's Alignment Project.

The programme's core products are its research outputs rather than commercial software. Research areas include mechanistic interpretability (sparse autoencoders, feature splitting and absorption), AI control (untrusted monitoring), activation probes for detecting deceptive reasoning, chain-of-thought monitoring, steganography detection in multi-agent LLM systems, and evaluation of frontier models. Notable outputs include a NeurIPS 2025 oral presentation on absorption in sparse autoencoders, a Best Paper Award at the Actionable Interpretability workshop at ICML 2025, and the open-source Model Organism Interpretability suite (54 model organisms across two base models) and EvalDetectBench. Publication rates exceed 80% at top venues, with 100% of the Summer 2024 cohort and 50% of the Spring 2025 cohort achieving NeurIPS acceptance.

LASR Labs is funded by grants from Arcadia Impact and Coefficient Giving, with additional grant support from UK AISI's Alignment Project. The programme provides £11,000 stipends to participants plus food, office space, and travel. The business model is non-commercial: the organization is a charitable initiative without equity ownership, and its 'revenue' consists of grant funding. The go-to-market approach is community-driven, recruiting through AI safety community networks (Alignment Forum, LessWrong), direct application via the website, and a dedicated Research Manager role paid £60,000–£90,000 per year. Customer segments are individual ML engineers and scientists seeking to transition into AI safety research careers, with alumni placing at UK AISI, Apollo Research, Leap Labs, and Open Philanthropy, or continuing into AI safety PhD programmes.

Name

LASR Labs

Legal name

LASR Labs

Website

https://lasrlabs.org

Company type

Private

Operating status

Operating

Headcount range

11–50 employees

Short description

LASR Labs is a 13-week AI safety research training programme in London that trains ML engineers and scientists to publish research on AI control and interpretability, funded by Arcadia Impact and grants, with alumni placed at UK AISI, Apollo Research, and Open Philanthropy.

Ownership category

akta.pro rank

Product category

AI Safety Research Training

NAICS

Scientific Research and Development Services (5417)

akta.pro primary industry

Safety & Alignment Evaluation (red-teaming, harmful capability testing) (HDAAAMAL)

akta.pro secondary industry

Safety, Alignment & Content Moderation (Guardrails, Red Teaming) (HDAAACAI)

Keywords

• Ai safety research
• Research training programme
• Mechanistic interpretability
• Ai control research
• Alignment research

Headquarters

HQ city

London

HQ country

United Kingdom

HQ region

Europe

Offices1 record

Markets served

GTM type

B2C

Offering type

Services

Cost components

Personnel, Operations, Infrastructure, Others

Revenue model

1. Programme Funding: LASR Labs operates as a research training programme funded by grants from Arcadia Impact (a registered charity) and Coefficient Giving. The programme provides £11,000 stipends to participants, food, office space, and travel as part of its core offering.
2. Research Output: The programme generates academic research papers that are published at top ML venues (NeurIPS, ICLR, ICML), building reputation and attracting further funding.

Pricing tiers

Go-to-market motion1 record

Distribution channels1 record

Marketing channels4 records

Core offering

LASR Labs runs a 13-week, full-time, in-person AI safety research training programme twice per year from its London offices. Participants work in teams of three to four, supervised by experienced AI safety researchers, to take a research project from proposal through to an academic-style paper and accompanying blog post. Participants receive an £11,000 stipend plus food, office space, and travel, with the goal of transitioning ML engineers and scientists into technical AI safety research careers.

Product overview

LASR Labs is not a software product company; it is a 13-week technical AI Safety research training programme that runs twice per year. Participants work in teams supervised by experienced AI safety researchers to write academic-style papers and accompanying blog posts. The programme does not offer commercial software products, SaaS offerings, or technology platforms for external customers.

Differentiator

Problem solved

Functional benefit

Quantifiable outcome

• 80%+ publication rate at top venues (NeurIPS, ICLR)
• +3 more outcomes

Named customers4 records

Segments2 records

Ideal customer profiles1 record

Technology focussed Yes

API detail

Has API

No

API docs

API detail

Core technology

AI maturity

App detail

Feature6 records

Partnerships

Three partnerships are on record, tiered core and minor.

• UK AI Security InstitutecoreStrategic or Co-development PartnerUK AISI provides supervisors for LASR Labs research projects. Charlie Griffin, who researches AI Control at UK AI Security Institute, acts as an advisor. Arcadia Impact recently secured a grant from UK AI Security Institute's Alignment Project to launch a year-long research programme working with their alignment team.
• London Initiative for Safe AI (LISA)coreStrategic or Co-development PartnerTeams work full-time and in person from the London Initiative for Safe AI (LISA). LISA provides the physical office space and infrastructure for the programme in London.
• Impact FirstminorOthersLASR Labs operates out of the Impact First offices in London, providing workspace for participants.

Scale indicators4 records

Recent moves6 records

Expansion highlights5 records

Direct peers

• MATS (ML Alignment & Theory Scholars): Operated by Apart Research, MATS is a stipend-funded AI alignment research training programme that matches participants with mentors to produce original research — the closest direct competitor to LASR in format, mission, and target participant profile.
• Constellation: Constellation is an AI safety research and training organisation working on alignment and governance, sharing LASR's mission of producing technical AI safety research and field-building.
• BlueDot Impact: UK-based organisation running AI safety courses and the AI Safety Fundamentals programme, sharing LASR's London geography, funder ecosystem, and mission of building the AI safety field — comparable as a parallel entry-point into the same talent pipeline.
• Apart Research: Apart Research runs MATS and other AI safety research programmes and events; it is a direct peer as both an operator of training cohorts and a producer of AI safety research outputs similar to LASR.
• ARENA (Alignment Research Engineer Accelerator): ARENA is a multi-month AI alignment training programme focused on building technical alignment engineering skills, directly comparable to LASR's transition-into-AI-safety mission and overlapping heavily on curriculum and supervisor pool.
• Apollo Research: Apollo Research is an AI safety evaluation organisation producing technical research on deceptive and scheming AI behaviour; it is both a downstream employer of LASR alumni and a producer of overlapping research (e.g., capability evaluations).
• AI Safety Camp: AI Safety Camp runs shorter, project-based cohorts for aspiring AI safety researchers, overlapping with LASR on mission and participant profile but with a lighter research-publication focus.

Others

• Future of Life Institute (FLI): FLI is a major AI safety grant-making and advocacy organisation that funnels philanthropic funding into the AI safety ecosystem in which LASR operates, making it a comparable ecosystem participant rather than a direct programme competitor.

Emerging players

• SaferAI: SaferAI is an emerging AI safety organisation focused on evaluating frontier AI developers and producing field-level assessments, partially overlapping LASR's capability evaluation research agenda.

Broad incumbents

• CHAI (Center for Human-Compatible AI, UC Berkeley): CHAI is an established academic research centre producing foundational AI alignment research; comparable as a producer of technical alignment work and a producer of aligned researchers via its PhD pipeline.

Market position

Strengths5 records

Weaknesses5 records

Competitive moat4 records

Key risks5 records

Key highlights6 records

Customer concentration

• LinkedInlinkedin.com/company/lasr-labs

Revenue estimate

Valuation estimate

Number of profiles

Profiles4 records

Funding overview

Funding rounds

Investors

Funding detail is available on the Subscription and Enterprise plan.Contact sales →

M&A

Investments

M&A and investment is available on the Subscription and Enterprise plan.Contact sales →

Frequently asked questions about LASR Labs