Major AI Companies Score Below Average On Safety Research Shows
AI industry has been championing self-regulation, but the results from the recently released independent report on AI safety don’t make an argument for its case. An independent AI safety watchdog, Life Institute, released their report card on AI safety in July 2026, and the results are abysmal, to say the least. The grading focused on areas such as risk assessment and management, transparency in doing it, and whether AI companies are honoring their own safety pledges.
The highest score was a C+ awarded to Anthropic AI despite its opting out of its commitment to guarantee safety before shipping its models. According to the report, OpenAI and Google’s DeepMind were awarded a C-Plain while Meta followed with a D+. At the bottom of the chain were xAI and China’s DeepSeek, which received a failure(F).
This should send alarm bells and should inform on what governments do going forward.AI is changing very quickly, and it is becoming clearer by the day that the industry cannot be fully trusted to be left on its own devices. I know we all don’t like governments getting into our businesses, but in this case, intervention by the government may prove to be a net positive.
We have already seen extreme cases whereby earlier this year, during the first day of the US-Iran war, AI was used for geolocation, in which a US missile strike wrongfully targeted a school, killing over 200 school kids in the process. We have also seen reports of fraud through voice cloning and even cases of AI driving individuals to commit suicide.
These are just a few cases of self-harm attributed to AI safety; it isn’t a theoretical issue anymore, as we have begun seeing real-life ramifications. I am all for self-policing, but if the best grade for AI safety is average, you can’t blame the critics for being skeptical.
