Anthropic warns AGI could arrive by 2028, says US must stay ahead of China in AI race

HIGHLIGHTS

Anthropic claims future AI systems may soon perform scientific and engineering tasks at expert human level.

The company warned that losing the AI race could allow China to expand AI use in surveillance, cyber operations and military systems.

Anthropic called for tighter chip export controls and restrictions on overseas compute access to preserve the US advantage in AI development.

Anthropic has published a new research and policy paper warning that artificial general intelligence (AGI) can come to reality by 2028 and urging the United States to maintain its technological lead over China in advanced AI development. Taking to paper titled 2028: Two Scenarios for Global AI Leadership, Anthropic argues that AI systems are advancing rapidly and may soon handle complex scientific, engineering, cybersecurity and research tasks at levels comparable to human experts.

The company described the future as having a country of geniuses in data centers, where AI models can accelerate scientific discovery, software development and even AI research itself. The company states that the nation leading advanced AI will also get major economic, political and military influence globally. The company specifically warned that authoritarian governments could potentially use powerful AI systems for mass surveillance, cyber operations and social control.

Also read: Did OpenAI share your ChatGPT chats with Google and Meta? New lawsuit raises concerns

According to the report, the United States currently has an advantage in AI due to its dominance in high-end chips and the computing infrastructure required for training frontier AI models. However, Anthropic stated that if loopholes in semiconductor exports, overseas compute access, and AI model access are not addressed, this lead may narrow.

The company also claimed that Chinese AI firms have remained competitive by leveraging overseas compute infrastructure and what it described as distillation attacks, in which advanced AI model outputs are allegedly used to recreate similar capabilities at a lower cost.

Anthropic has mentioned two potential outcomes for 2028. In the first scenario, the United States and its allies tighten export controls, preventing chip smuggling and accelerating domestic AI adoption. According to the company, this could allow democratic nations to maintain an AI capability advantage of 12 to 24 months over Chinese rivals.

In the second scenario, lax enforcement and continued access to overseas infrastructure allow China to stay close to the AI frontier. Anthropic warned that this would allow China to aggressively shape the global AI ecosystem.

Ashish Singh

Ashish Singh is the Chief Copy Editor at Digit. He's been wrangling tech jargon since 2020 (Times Internet, Jagran English '22). When not policing commas, he's likely fueling his gadget habit with coffee, strategising his next virtual race, or plotting a road trip to test the latest in-car tech. He speaks fluent Geek.

Connect On :