Berkeley Ai Safety Student Initiative

What Berkeley Ai Safety Student Initiative does

Berkeley AI Safety Student Initiative (BASIS, formerly Berkeley AI Safety Initiative) is a student-run educational organization at UC Berkeley that operates a portfolio of free programs aimed at students interested in technical AI safety research and AI policy. Its core offerings include the Technical AI Safety Decal (a semester-long weekly course held Wednesdays 6-8 PM in Dwinelle 88 covering RLHF, adversarial robustness, LLM agent safety, reward hacking, interpretability, and representation engineering), an AI Policy Fellowship with weekend meetings on governance frameworks, a weekly AI Policy Reading Group open to both Berkeley students and outside community members, and an ongoing Speaker Series featuring researchers from OpenAI, DeepMind, and Anthropic. Membership acquisition runs through Slack, an Airtable-hosted interest form and newsletter, and on-campus infosessions; all programs are offered free of charge.

The organization has no commercial revenue model. It does not sell products, license technology, or charge membership fees; its sole operating expense indicators visible from the data are a Squarespace-hosted website and a personal Gmail contact address ([email protected]). The organization is led by a student leadership board of 15+ members, including co-presidents Vinaya Sivakumar (cognitive science and technology policy) and Dylan Xu (CS, fourth-year), program directors, and an advisor (Tzu Kit Chan) with prior operations experience at Atlas Fellowship and the ML Alignment & Theory Scholars Program. A board member is a Berkeley Law LL.M. candidate with a decade of M&A, securities, and fintech legal practice.

BASIS members participate in cross-university AI safety research published on arXiv, with co-authors from UC Berkeley, the University of Toronto, MIT, the University of Cambridge, the University of Waterloo, and Georgia Tech; outputs include the WMDP Benchmark, HarmBench, MASK Benchmark, Remote Labor Index, Representation Engineering work, and Humanity's Last Exam. The organization is incorporated informally as a UC Berkeley student group with no disclosed parent company, venture funding, or formal corporate entity, and its previous name "Berkeley AI Safety Initiative" has been retired in favor of the current designation.

Name

Berkeley Ai Safety Student Initiative

Legal name

Berkeley AI Safety Student Initiative

Website

https://berkeleyaisafety.com

Company type

Private

Operating status

Operating

Short description

Berkeley AI Safety Student Initiative (BASIS) is a UC Berkeley student organization offering free technical AI safety coursework, AI policy fellowships, reading groups, and speaker events featuring OpenAI, DeepMind, and Anthropic researchers, serving Berkeley students and broader community members interested in AI safety and governance.

Ownership category

akta.pro rank

Product category

Educational Services and Student Research Organization

NAICS

Educational Support Services (611710)

akta.pro primary industry

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

akta.pro secondary industries

Safety, Alignment & Content Moderation (Guardrails, Red Teaming) (HDAAACAI), Responsible AI, AI Governance & Compliance Services (BPAEAHAJ)

Keywords

• AI safety education
• Student research programs
• AI policy fellowship
• Technical AI safety
• AI alignment curriculum

GTM type

B2C

Offering type

Services

Cost components

Personnel, Operations, Marketing or Sales

Go-to-market motion1 record

Distribution channels4 records

Marketing channels4 records

Core offering

Berkeley AI Safety Student Initiative (BASIS) is a UC Berkeley student organization that delivers educational programming on AI safety, including a Technical AI Safety Decal covering topics such as RLHF, adversarial robustness, LLM agent safety, reward hacking, interpretability, and representation engineering, plus an AI Policy Fellowship, an AI Policy Reading Group, and a Speaker Series featuring researchers from OpenAI, DeepMind, and Anthropic. Programs are offered free of charge to UC Berkeley students and the broader community.

Product overview

Berkeley AI Safety Student Initiative (BASIS) is a student organization offering a portfolio of educational programs: the Technical AI Safety Decal provides technical coursework on AI safety topics; the AI Policy Fellowship and Reading Group offer governance and policy-focused discussions; the Speaker Series hosts industry experts; and members conduct research on AI safety evaluation and alignment. The organization functions as an educational and research collective rather than a commercial product company.

Differentiator

Problem solved

Functional benefit

Products and services

• Technical AI Safety Decal
• AI Policy Fellowship
• AI Policy Reading Group
• Speaker Series
• AI Safety Research Projects

Segments3 records

Ideal customer profiles3 records

Technology focussed No

API detail

Has API

No

API docs

API detail

Core technology

AI maturity

App detail

Partnerships

Four partnerships are on record, tiered minor and core.

• OpenAIminorOthersOpenAI researchers and professionals serve as guest speakers in BASIS Speaker Series events, providing students with insights into frontier AI safety work. This is an informal speaker relationship rather than a formal partnership.
• DeepMindminorOthersDeepMind researchers and professionals serve as guest speakers in BASIS Speaker Series events, sharing technical AI safety research perspectives with student members.
• AnthropicminorOthersAnthropic researchers and professionals serve as guest speakers in BASIS Speaker Series events, providing student exposure to AI safety and alignment work at Anthropic.
• Berkeley AI Safety Initiative (BASIS) Members (Various Collaborators)coreOthersBASIS members collaborate on AI safety research projects published on arXiv, working alongside researchers from UC Berkeley, University of Toronto, MIT, University of Cambridge, University of Waterloo, Georgia Tech, and other institutions. Research projects include Remote Labor Index, The WMDP Benchmark, HarmBench, and Humanity's Last Exam.

Scale indicators2 records

Recent moves6 records

Expansion highlights5 records

Broad incumbents

• Anthropic: Frontier AI lab whose researchers serve as BASIS speakers and whose fellowship programs influenced BASIS's curriculum. Tangentially related as a target employer and content source, not a direct competitor.
• Center for AI Safety (CAIS): Larger, established nonprofit focused on AI safety research, advocacy, and education. Comparable mission but operates at significantly larger scale with full-time staff, formal partnerships, and broader educational reach beyond any single university.
• 80,000 Hours: Established nonprofit providing career guidance for high-impact roles including AI safety and policy. Comparable mission of directing talent into AI safety careers, but operates at much larger scale with paid staff and broader scope.

Direct peers

• AI Safety Camp: International student-oriented program running short, intensive AI safety research camps producing arXiv papers. Comparable as an educational research collective aimed at early-career researchers in alignment and safety.
• Stanford AI Safety / Stanford Existential Risks Initiative: Stanford's analogous student-led AI safety community offering educational programming, speaker events, and policy/technical tracks for Stanford students. Most direct comparable given similar university setting, scope, and dual technical/policy focus.
• MATS (ML Alignment & Theory Scholars): Research fellowship program placing students with AI safety mentors for technical alignment research. Comparable as an early-career AI safety talent program; a BASIS advisor previously ran operations at MATS.
• Apart Research: AI safety research organization running fellowships, hackathons, and collaboration events that produce arXiv publications similar to BASIS's research output. Comparable as a student-facing research collective focused on alignment and evaluation.
• MIT AI Alignment Group: MIT student organization focused on AI alignment research and education, running reading groups, technical seminars, and research projects. Directly comparable as a peer university student org with similar technical alignment focus.
• BlueDot Impact: UK-based nonprofit running the AI Safety Fundamentals course and AI governance programs that explicitly informed BASIS's curriculum. Directly comparable as an educational organization building AI safety talent pipelines through cohort-based courses.
• Cambridge AI Safety Hub: Cambridge University's student AI safety community offering technical and policy programming, speaker events, and research projects. Direct peer as a UK counterpart with similar educational model and speaker series.

Market position

Strengths4 records

Weaknesses4 records

Competitive moat5 records

Key risks5 records

Key highlights6 records

Customer concentration

Revenue estimate

Valuation estimate

Number of profiles

Profiles14 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 Berkeley Ai Safety Student Initiative