Anthropic CEO calls for slower AI race as safety concerns intensify
Amodei
said in a post that AI could, within six to 12 months, become capable of
directing large networks of autonomous agents across the internet. He argued
that even a one- or two-year delay in reaching more advanced capabilities could
provide valuable time to strengthen safeguards and alignment research.
His
comments come amid growing concern among AI researchers and employees over the
risks associated with increasingly autonomous systems. Several industry workers
have recently resigned, citing concerns that companies are prioritizing
competition in the race toward advanced AI.
Amodei
proposed that major AI developers give independent safety teams ongoing access
to their operations so researchers can monitor how frontier systems are
developed and tested. Anthropic plans to adopt the approach internally, he
said.
OpenAI
CEO Sam Altman said his company would also commit to the independent evaluator
proposal and said the company was prioritizing safety work before considering a
stock market debut.
Amodei
acknowledged that stronger safeguards would require cooperation among
technology companies and governments, including coordination across countries
with different political systems.
Despite
his concerns, he said AI could deliver major benefits, including advances in
medicine, if its development remains subject to effective oversight and
control.
