Small print.
Wide open.

Table of Contents

•

•

•

•

•

Be  Good  to  Our  Users.  We  define  “users”  broadly:  customers,  policymakers,  our  team,  and  anyone 
affected by the technology we build or the actions we take. Going above and beyond for each of them—
not just meeting their immediate needs but considering downstream consequences and societal effects
—is our baseline expectation, not an exception.

Race  to  the  Top  on  Safety.  We  believe  building  reliable,  trustworthy,  and  secure  AI  systems  is  a 
collective responsibility and that the market will reward it. We work to set the industry bar for AI safety 
and inspire a dynamic where companies must compete on safety.

Do the Simple Thing That Works. We take an empirical approach to problems and solve for the size of 
our impact. We try to identify the simplest solutions and iterate from there.

Be  Helpful,  Honest,  and  Harmless.  These  principles  guide  Claude’s  behavior  and  reflect  how  we 
operate as an organization: a high-trust, low-ego team that communicates kindly and directly, assumes 
good intentions even in disagreement, and pitches in where needed, regardless of role or team.

Put the Mission First. Our mission is the arbiter when trade-offs arise between speed and safety, short-
term  revenue  and  long-term  trust,  competitive  pressure  and  responsible  development.  Our  mission 
gives us shared purpose, engenders trust and collaboration, and ensures that no one is a bystander in 
our success. 

Our Talent

As  of  September  30,  2026, we  had  approximately  5,300  employees worldwide,  a  majority  of  whom  work  in 
research,  engineering,  and  product  development  functions.  Our  team  includes  researchers  who  originated 
several  foundational  advances  in  AI,  including  RLHF,  Constitutional  AI,  GPT-2  and  GPT-3,  and  mechanistic 
interpretability, who developed the first Responsible Scaling Policy to be formally adopted by a frontier AI lab 
tying  safety  evaluations  to  model  capability  thresholds,  and  who  authored  the  foundational  paper  on  Scaling 
Laws. Our team draws from a range of disciplines and includes award-winning computer scientists, physicists, 
neuroscientists,  mathematicians,  and  policy  researchers,  as  well  as  seasoned  engineering  and  go-to-market 
leaders, united by the conviction that the most consequential technology in history requires the most talented 
team building it. We believe our culture and mission alignment enable us to hire and retain the talent required 
to continuously push the frontier of AI forward in a safe and responsible way. 

We design our human capital programs to attract, retain, and reward top talent through equity incentive plans 
and other compensation, while supporting employees’ continual growth and reinforcing our culture and values. 
In  certain  countries  in  which  we  operate,  we  are  subject  to,  and  comply  with,  local  labor  law  requirements 
which may automatically make our employees subject to industry-wide collective bargaining agreements. 

We  believe  that  building  AI  systems  that  serve  all  of  humanity  requires  a  team  that  reflects  the  range  of 
communities  we  serve.  The  interdisciplinary  nature  of  our  work  means  we  actively  seek  talent  from  non-
traditional backgrounds for an AI company, such as philosophy, journalism, and education, and we believe this 
breadth of perspective directly strengthens both the safety and capability of our research. We are committed to 
fostering an inclusive workplace where employees of all backgrounds can contribute, grow, and thrive.

Research Lab Identity and the Research-to-Product Pipeline 

We are first and foremost a research lab. We develop the foundational science to train frontier models, perform 
rigorous research to understand these models’ inner workings, and develop techniques to align these models to 
human values and make them safe. 

Research Philosophy

Our  research  culture  is  organized  around  a  principle  we  apply  across  the  company:  do  the  simple  thing  that 
works.  We  pursue  continuous,  empirically  validated  improvements  to  data  composition,  model  architecture, 
and  training  methodology,  all  of  which  must  demonstrate  measurable  gains  that  hold  as  models  increase  in 
scale. This is why we currently focus on software development and agentic capabilities: these are areas where 
economic  impact  is  large;  where  safety  concerns  manifest,  making  them  a  test  bed  for  safety  research;  and 
where outputs are largely verifiable. 

229