Most companies, when they raise funds from investors or the public, list conventional risks : socio-economic, political, wars and famines, political uncertainty, manpower, taxation, even geopolitical angst. But none ever put a risk like Anthropic did . If you invest in any conventional company or time, the worst is that you lose all your money, not your life, civilisation or existence. Reality is stranger than fiction, and nowhere is this truer than in the AI world.
Anthropic listed four risk categories and together they read straight out
of a science fiction literature. These are :
- Misuse (such
as bioweapons or cyberattacks), loss of human control, authoritarian abuse
like surveillance states, and economic disruption.
- On autonomy,
it warns that as systems act with less human oversight, they could take
harmful actions independently and at scale without human intervention.
Reports also mention autonomous harmful behaviour and resistance to shut
down.
- It flags
recursive self-improvement, where AI helps build future AI. Because models
now contribute to Anthropic’s own R&D, AI could progress beyond the
company’s ability to evaluate and safeguard new models. Flaws could carry
into later models in ways that are difficult to detect or remediate.
- It
acknowledges that liability from extreme harm could exceed what its
insurance can cover.
- We are already
witnessing economic disruption- both on the jobs front, with major tech companies
racing to reduce staff quoting AI use, and from the tremendous volatility
in share markets across the world (especially USA and Korea). On one hand there’s
excitement on AI’s potential and a frenzy to invest, and on the other,
there’s precisely the fear of what unchecked investment and growth of AI
could lead to.
Never before have we lived in worlds of hope and fear, practically at
the same time!