Just say no to OpenAI and Anthropic

In his recent book End Times Fascism and the Fight for the Living World, Naomi Klein and Astra Taylor highlight the alarming influence of tech oligarchs in the rise of artificial intelligence. As a software and security engineer, one aspect that particularly struck me is the belief held by the creators of leading language models that they are creating a "Machine God," and that they are willing to risk the world to achieve artificial general intelligence (AGI).

This sentiment, widely expressed in the AI hype surrounding OpenAI and Anthropic, is not unique to these two companies. In fact, open weight models, which are as powerful as the latest ChatGPT and Claude models, are readily available. Unlike the closed versions offered by OpenAI and Anthropic, open weight models provide significant advantages.

They are not limited by corporate restrictions and do not require users to agree to terms that grant the companies access to their data. However, while these models offer greater privacy, they still have strict guardrails that monitor user sessions for potential misuse. Some users may gain access to less restricted versions of these models, but this access is often limited to certain groups, such as cybersecurity companies and US military contractors.

The companies themselves, such as OpenAI and Anthropic, have the final say in who gets access to their more capable, less policed models. This undemocratic approach has raised concerns about the concentration of power in the hands of a few individuals. Instead of relying on these corporate overlords, it is suggested to explore alternative options, such as open weight models like GLM-5.3 and GLM-5.3-Flash.

These models, while not perfect, provide a safer and more secure alternative for those seeking to leverage AI technology without compromising their privacy. Other open weight models from Chinese companies, such as DeepSeek, Qwen, and Kimi, are also worth considering. While these models are impressive, it is crucial to remember that trust should not be placed in the companies themselves, as they may share user data with the Chinese government.

Mistral, a French frontier AI lab, stands out as a more trustworthy option due to its independent nature. Secure enclaves and remote attestation are emerging technologies that can provide a meaningful difference in privacy compared to the current status quo. Companies like Tinfoil and Maple AI are actively working on implementing these solutions, offering varying levels of accessibility and pricing.

For those seeking a ChatGPT-like experience, Maple AI is recommended, with its consumer-focused approach and affordable subscription plans.