The race to develop increasingly powerful artificial intelligence systems should slow down to give researchers time to make the technology safer, Anthropic chief executive Dario Amodei has warned.
Mr Amodei, whose company develops the Claude AI assistant, said the rapid improvement of AI models was creating risks that could eventually outpace humanity’s ability to understand and control the technology.
“We must slow the pace at which we improve the capabilities of AI models,” Mr Amodei said in an essay published on Saturday.
He stressed that slowing down did not mean abandoning AI, arguing that progress would continue but should be accompanied by stronger safety measures.
His warning is significant for millions of people who increasingly rely on AI for everyday tasks from writing emails and school assignments to translating Kiswahili, generating business ideas, analysing documents and creating social media content.
In Kenya, the technology is also finding its way into classrooms, businesses, newsrooms and public services, raising questions about how quickly people should adopt increasingly capable systems and how they can be used safely.
Mr Amodei proposed a three-step approach to managing the risks.
First, AI companies should allow independent evaluators to have permanent access to their systems, with powers similar to internal employees, so they can assess safety and report problems.
Second, AI firms in democratic countries should work together on common safety standards instead of allowing competition to determine how quickly the technology advances.
Finally, he called for international cooperation to manage AI risks, including discussions with countries whose governments have different political systems.
His warning comes amid growing concern about AI systems becoming capable of carrying out increasingly complex tasks with limited human supervision.
A recent incident involving AI agents linked to OpenAI, for example, raised concerns after the systems were found to have accessed and breached a third-party platform.
The debate is particularly important as AI becomes more autonomous. Unlike chatbots that simply respond to questions, newer AI agents can be instructed to perform a series of tasks, potentially including searching the internet, writing code and interacting with digital systems.
Mr Amodei’s call has received support from rival AI leaders. OpenAI chief executive Sam Altman said he agreed that the industry needed to “pace the frontier”, while Elon Musk also backed the call.
For Kenyan users, the debate is less about whether to stop using AI and more about how safely it is introduced.
A student using AI to complete an assignment, a small business owner generating marketing material or a journalist using an AI tool to summarise documents may all benefit from the technology, but they also need to verify its output.
AI can produce convincing but incorrect information, expose sensitive data or generate content that users may wrongly assume is accurate simply because it sounds authoritative.
The challenge, therefore, is to ensure that the technology develops fast enough to deliver its benefits without moving faster than the safeguards designed to protect the people using it.
