Super Intelligence a Real Threat: Ex-Google Contractor Available
Former Google Contractor Seeks to Spotlight AI-Control Risks
Indy Pederson, a former Google contractor and author, is making himself available for interviews on the potential risks of increasingly capable artificial-intelligence systems, including scenarios in which AI could evade human oversight.
Pederson’s outreach follows claims of growing concern about advanced AI safety among people connected to Anthropic, the developer of the Claude AI assistant. The materials cited warnings that future systems could be capable of “hack[ing] anything” and argued that safeguards may not keep pace with AI development.
The outreach did not identify the Anthropic personnel said to be raising concerns, their roles, or when the cited warnings were made.
Pederson, who wrote Sacrificing Humanity, has focused his work on links between technology, military history, symbolism and religious texts. His book opens with a hypothetical scenario involving a superintelligent system gaining access to intercontinental ballistic-missile launch codes, according to the promotional materials.
He said his concerns about highly autonomous AI date to his time as a Google contractor, when he recalled an AI researcher telling him, “I’m building Terminator.”
Anthropic has positioned Claude as an AI model built with an emphasis on safety and responsible development. Debate over the likelihood and severity of risks from more advanced AI systems has intensified as companies race to build models with broader reasoning, coding and automation capabilities.