Bruce Schneier and Nathan E Sanders argue that, if OpenAI and Anthropic fail to meet investors’ expectations, the US government should nationalise them and operate them as democratically controlled national laboratories. They say this could restore the firms’ original public-interest aims and prevent powerful AI systems from being governed chiefly by corporate profit motives.

The authors contend that frontier AI is a weak long-term business because models are costly to train, quickly become outdated and increasingly resemble commodities. They cite pressure from open-source and Chinese competitors, reduced enterprise spending on AI usage, opposition to datacentres and uncertainty over profitability, despite recent IPO enthusiasm and talk of trillion-dollar valuations.