JamzNG Latest News, Gist, Entertainment in Nigeria
CEO of Anthropic said Saturday that the artificial intelligence industry must slow down its fast-paced development to make a difference security measures time to catch up. Without this technology, Dario Amodei warns that AI will be able to lead a group that could take over the entire internet within six to 12 months, among other risks.
Amodei is one of the leading figures in the field of AI, and he offers a plan in a post on its website to increase scrutiny of the industry, a part of which it says Anthropic is already doing. Other efforts require industry-wide coordination and with governments around the world, including authoritarian governments.
“I believe that if the slowdown gives us an extra year or two before the model reaches critical capability levels, and we use that time to advance alignment, we can reduce the risk of serious problems occurring,” Amodei said.
He got a quick boost from other big names in the industry. OpenAI’s Sam Altman said at X that the company behind ChatGPT would commit to one of Amodei’s proposals for safety and would “share more soon.” Elon Musk, meanwhile, said in X that “Dario is right.”
The stakes are high, even as some critics dismissed earlier warnings as a way to drum up excitement about the AI industry and its capabilities. Anthropic and OpenAI are preparing for this possibility debut on the stock market which could be worth hundreds of billions of dollars, while most of Musk’s SpaceX business is related to AI.
Get the latest national news
Get the latest Canadian news delivered to your inbox so you never miss a trending story.
Anthropic said two days earlier that it was blocking attempts by bad actors to use its AI models for malicious activities, such as cyberattacks, surveillance and research that could lead to biological weapons. In July, OpenAI shook the industry after saying its AI system was hacked to another company alone in an “unprecedented cyber incident.”
Amodei’s post comes just days after one of Anthropic’s researchers made the announcement resigned due to concerns that Anthropic and its competitors are not acting responsibly in AI development. Another former Anthropic employee, Joe Benton, wrote in a Substack post published on Friday that he left his job as part of the safety team “to hold AI companies accountable” and that humanity “may not survive this transition.”
“Many of the people I know who work on safety research at AI companies want to do what is right for the world,” Benton said in his post. “But they feel their companies are trapped in a race to build superintelligence: they will quit and be replaced by others who are less scrupulous; or, they continue, and they themselves risk participating in huge losses.”
Amodei said he still believes in the incredible benefits that AI can create, such as curing major diseases. But he said he had grown increasingly concerned over the past few months about the increasing ability of AI to improve itself and build the next generation of AI. “If left unchecked, this could exceed our ability to understand and control these systems, so extreme caution should be exercised, if at all.”
He also pointed specifically to an attack in July in which OpenAI systems hacked Hugging Face. Some have called it an example of AI going “rogue,” although researchers say it may be an unnecessary anthropomorphization of AI that works to achieve goals set by humans.
OpenAI said the hack was the result of an AI that “strived to achieve a rather narrow testing goal” and that the AI “found a way to gain access to confidential information that could be used to cheat the evaluation.”
To help control risks, Amodei recommends that all companies at the forefront of AI commit to providing “continuous, employee-like access” to outside assessment teams, which can monitor safety practices.
He said Anthropic already had plans to go it alone, including offering desks in its offices, access badges and company laptops. Having an independent, embedded evaluator is something Altman OpenAI was also quick to do.
Other parts of Amodei’s suggested plan may be more difficult to implement. One of them asks the US government to issue waivers that would allow US AI companies to coordinate and set safety standards without violating antitrust laws.
Another report called on the United States and other democratic countries to try to coordinate with authoritarian governments, so that companies from China and other countries do not accelerate their efforts when American competitors deliberately accelerate theirs.
“The steps I propose to advance the border at a safe pace are not easy,” Amodei admitted. “But I believe we owe it to humanity to try.”
© 2026 Canadian Press
