Open AI Models Gaining Power—But Safety Standards Falling Behind

Powerful Open AI Models Are Racing Ahead—But Safety Standards Aren't Keeping Up

A new SaferAI report has found that Z.ai's open-weight GLM-5.2 model is approaching the capabilities of frontier AI systems like ChatGPT and Claude, but it's missing critical safety protections that companies use to prevent misuse. The discovery is renewing concerns that open-source AI—which anyone can download and modify—could become powerful enough to cause real harm before safeguards catch up.

For small business owners, this matters because you're increasingly relying on AI tools to handle customer data, financial information, and business strategy. As more powerful models become freely available and easy to customize, the risk that someone uses AI to attack your business, steal customer information, or create convincing fake content grows. Right now, the companies building frontier AI have teams focused on safety. Open models don't always have that same oversight.

The gap between capability and safety is widening. Frontier models—the ones made by OpenAI, Anthropic, and Google—go through months of testing to ensure they won't help with illegal activities or produce harmful content. Open models can skip those steps entirely. That means a business competitor, bad actor, or even a disgruntled employee could theoretically use a more powerful open model with fewer guardrails to target your company.

This doesn't mean you should panic or stop using AI tools. But it's a sign that the AI landscape is becoming more fragmented. The tools you trust from major companies may have better safety records than alternatives, and that's worth considering when you're evaluating which platforms to use with sensitive business information. Not relying on a single AI tool is smart strategy anyway—having multiple providers means you're not vulnerable if one gets compromised.

What to watch: Keep an eye on whether governments start setting minimum safety standards for open-source AI, and whether major model providers release more transparent information about their safety testing. That transparency could help you make better decisions about which tools are safe for your business.