Small print.
Wide open.

Table of Contents

our industry can be independently verified. We have further supported regulation and oversight of frontier AI, 
including  measures  focused  on  transparency  and  independent  third-party  auditing,  and  international 
coordination,  so  these  practices  apply  across  frontier  developers  rather  than  depend  on  voluntary  adoption 
alone.

We  believe  frontier  AI  companies  and  governments  should  take  action  to  strengthen  the  security  of  critical 
infrastructure against AI-related and other cyber threats, including by:

•

•

•

•

•

•

providing  responsible  model  access,  funding,  training,  and  hands-on  support,  particularly  for  under-
resourced critical-infrastructure defenders; 

building observability and security tools; 

working to ensure agentic identities are traceable and accountable; 

sharing best practices in continuous monitoring; 

investing in authorized testing, private disclosure, and verified remediation of vulnerabilities; and

sharing tools, playbooks, and threat assessments with governments, security partners, and open-source 
maintainers. 

We believe that by fixing the most dangerous weaknesses, verifying the fixes, and sharing what works, frontier 
AI companies and governments can turn today’s AI advances into lasting improvements in security that benefit 
everyone. We have committed up to $100 million in cyberdefense-related initiatives through Project Glasswing, 
and  we  expect  to  devote  significant  additional  resources  over  time  toward  strengthening  the  security  of  AI 
systems and the critical infrastructure that depends on them. The scope, timing, and amount of any such future 
investments  have  not  been  finalized  and  remain  subject  to  change  based  on  business,  regulatory,  and  other 
considerations, and there can be no assurance that these initiatives will achieve their intended effect.

Our  own  trajectory  demonstrates  that  commercial  success  and  a  serious  commitment  to  safety  are  not  in 
tension.  We  have  built  one  of  the  fastest-growing  businesses  in  history  while  maintaining  safety  as  a  core 
priority at every stage. 

Our Commitment to Public Benefit

We  founded  Anthropic  on  the  belief  that  the  most  transformative  technology  in  human  history  demands  an 
extraordinary  level  of  responsibility,  and  that  responsibility  extends  far  beyond  building  safe  systems.  Our 
public  benefit  purpose  is  the  responsible  development  and  maintenance  of  advanced  AI  for  the  long-term 
benefit  of  humanity—a  mission  in  which  safety  is  central  to  our  work.  Powerful  AI  can  deliver  immense 
benefits, but only if it is developed and deployed with rigorous attention to its risks. At times, that may require 
pausing the launch of our own most capable systems if we conclude they cannot yet be deployed safely. It also 
means  actively  working  to  ensure  that  the  benefits  of  AI  reach  everyone,  that  the  organizations  and 
communities people depend on are strengthened, and that the hardest problems facing humanity are not left to 
market forces alone to solve.

Realizing  this  mission  requires  understanding  how  AI  is  actually  being  used  and  how  it  is  reshaping  society. 
Our  Societal  Impacts  team  builds  privacy-preserving  tools  to  study  real-world,  large-scale  uses  of  Claude—
examining  how  students,  educators,  and  workers  use  the  model,  how  its  values  surface  in  everyday 
conversations,  and  what  users  hope  and  fear  about  the  future  of  AI.  This  includes  what  we  believe  to  be  the 
largest multilingual qualitative study of its kind, in which more than 80,000 people shared in depth how they 
see AI shaping their lives. Our Economics team extends this work to the labor market, publishing recurring data 
reports and fully open datasets (the Anthropic Economic Index) on how Claude is used across jobs, industries, 
and regions, as well as modeling potential future economic scenarios, so that workers, employers, policymakers, 
and  independent  researchers  can  ground  their  decisions  in  evidence  and  rigorous  modeling.  To  deepen  and 
broaden  the  reach  of  this  work,  we  have  established  the  Anthropic  Institute,  an  effort  to  share  research  and 
internal  data  from  across  Anthropic  with  external  partners  and  the  public,  working  together  to  address  the 
biggest questions posed by Powerful AI.

234