Warning of "catastrophic harm": artificial intelligence is no natural disaster
TL;DR
Anthropic and OpenAI are once again alarmed by the pace of their own work, with Anthropic CEO Dario Amodei warning that an uncontrolled AI system could take over the internet within six to twelve months. In a commentary for netzpolitik. org, AlgorithmWatch co-founder Matthias Spielkamp pushes back: AI is no natural disaster, it is built and run by people. The real loss of control, he argues, comes when media and politicians amplify the companies' warnings uncritically and skip binding rules.
Nauti's Take
The commentary offers a valuable opportunity to ground the debate, reminding readers that AI systems only use the access and resources people give them, which keeps responsibility traceable. The limit is that even a well-meant reassurance can play down real risks such as agents with far-reaching permissions.
Anyone deploying AI agents at work should take exactly that away: grant permissions narrowly and define clearly who is liable.
Sources