Anthropic is advancing its artificial intelligence systems roughly twice as fast as a year ago, according to an independent analysis released Thursday, even as CEO Dario Amodei has urged the industry to slow down so safety measures can catch up.
The Berkeley, California-based nonprofit Model Evaluation and Threat Research, or METR, reviewed Anthropic’s internal data and interviewed staff. The assessment comes as the San Francisco company prepares an IPO later this month that could be the largest ever, amid debate over whether advanced AI could escape human control.
METR estimated that Anthropic is improving its models’ abilities on “well specified tasks” about twice as quickly as a year earlier. Researchers said much of the acceleration appears driven by AI itself; Anthropic says its Claude chatbot is increasingly helping develop more advanced versions.
The findings matter because most details about AI capabilities and risks have come from AI companies, which President Donald Trump’s administration has encouraged to police themselves on safety. During METR’s review, Amodei’s Sept. 12 essay urged the industry to “slow the pace at which we improve the capabilities of AI models.”
Anthropic co-founder Sam McCandlish said on X that many METR conclusions matched his own. “AI is speeding up our research, but it’s hard to say exactly how much,” he wrote. He said Anthropic would release its own analysis soon and that the METR work began before and is separate from its embedded third-party evaluator commitment.
One METR report drew on about 11 weeks of internal data from mid-August through late October plus employee interviews. METR said it had “unprecedented access” to Anthropic’s R&D process, though redactions limited detailed estimates; it still believed the rate of improvement had doubled.
“Anthropic appears to have made as much progress on benchmarks in the past year as they did between October 2023 and October 2025,” the report said. McCandlish, also Anthropic’s responsible scaling officer, wrote that acceleration has major safety implications, but outside observers and policymakers struggle to gauge it because most research happens inside frontier labs.
METR reviewed Anthropic’s public August risk report and a private version, saying the public document “overall understates the level of acceleration that was occurring.” McCandlish acknowledged redactions make it harder to check METR’s work but said revealing what drives the speedup is commercially sensitive and could give others a speedup roadmap.
The report included caveats: gains may be limited to easily graded tasks rather than messier ones, or earlier progress may have been understated. METR said it could confidently rule out increased computing power as the main cause and guessed that AI improving AI was driving the acceleration.
A key concern is that recursive self-improvement could compound acceleration, with no apparent compute or data bottlenecks. METR said AI cannot yet make fully autonomous contributions to key training stages, but Anthropic researchers described “signs of life” that it could happen soon. Amodei has warned unchecked self-improvement could outrun humans’ ability to understand and control AI.
Original Source: https://www.ndtv.com/world-news/anthropic-rapidly-accelerating-ai-development-amid-slowdown-calls-reports-12290126#publisher=newsstand
Category:
Tags:
Publish Date: 2026-11-06 02:31:00
