According to a major AI Safety Group, artificial intelligence companies are “fundamentally unprepared” for the results of creating systems with human-level intellectual performance.
The Future of Life Institute (FLI) said none of the AI Safety Index companies scored higher than the D in the “Existential Safety Plan.”
One of the five reviewers of the FLI report aims to develop artificial general information (AGI), but none of the scrutinized companies have “something like a consistent, practical plan,” making sure the system remains safe and controllable.
AGI refers to the theoretical stages of AI development in which a system can match humans when performing intellectual tasks. Openai, developer of ChatGpt, says that AGI is about ensuring that it “benefits all humanity.” Safety campaigners warn that AGIs can pose existential threats by avoiding human control and causing catastrophic events.
The FLI report states: “The industry is fundamentally unprepared for its own goals. Companies claim to achieve artificial general information (AGI) within 10 years, but existential safety plans do not score more than D. ”
The index evaluates seven AI developers, Google Deepmind, Openai, Anthropic, Meta, Xai, Zipuu AI and Deepseek, in six areas, including “current harm” and “existential safety.”
Humanity received the highest overall safety score in C+, followed by Openai and C- Google Deepmind.
FLI is a US-based nonprofit that campaigns to make cutting-edge technology safer and can operate independently for “unconditional” donations from Crypto entrepreneur Vitalik Buterin.
Saferai, another safety-focused nonprofit, has released a report on warnings that advanced AI companies have “very weak to weak risk management practices” and label current approaches as “unacceptable.”
FLI's safety grades were assigned and reviewed by a panel of AI experts, including UK computer scientist Stuart Russell and Sneha Revanur, founder of the AI Regulation Campaign Group.
Max Tegmark, co-founder of FLI and professor at the Massachusetts Institute of Technology, said it was “pretty harsh” to aim for cutting-edge AI companies to build ultra-intelligent systems without revealing plans to address the outcome.
He states:
Tegmark said the technology exceeds expectations, citing a previously thought belief that experts will have decades in addressing the challenges of AGI. “Now, the companies themselves say it's a few years away,” he said.
He shows that advances in AI capabilities have all shown improvements to their ancestors since the Global AI Summit in Paris in February, new models such as Xai's Grok 4, Google's Gemini 2.5 and its video generator Veo3 have all shown improvements to their ancestors.
A Google Deepmind spokesperson said the report does not take into account “all of Google Deepmind's AI safety efforts.” They added: “AI's comprehensive approach to safety and security is far beyond what's captured.”
Openai, Anthropic, Meta, Xai, Zhipu Ai and Deepseek have also been approached for comment.
