September 2026 AI news roundup: Google updates, LLMjacking, publisher payments, Muse and protein watermarking

Google’s latest AI updates were highlighted in AI news announced in September 2026, while related coverage pointed to security, monetization, assistant, and biology issues around AI.

Google analysts warned that stolen AI credentials are being sold underground, and businesses are footing the bill when those stolen credentials are used. The problem is described in the headline as LLMjacking, a term tied to the abuse of AI services. As a general matter, unauthorized access to an AI account can cause unexpected charges for the legitimate business.

Google’s shift to AI-powered search has also changed how the world’s largest search engine generates web traffic, rankling large and small publishers. The company has experimented in recent months with paying publishers when their content appears in AI answers, but a new report says the early attempt is struggling. AI answers can synthesize information from multiple websites, which can reduce the traffic that publishers previously received from search results.

Meta launched a new Muse AI agent that it claims can help users with tasks from firing off emails to buying stuff online. Coverage says Muse can be surprisingly effective if users are willing to trust Meta with their data and hand the agent their credit card. In general, AI agents are software helpers that act on a user’s behalf, so they often require broad access to personal information and payment methods.

The security conversation extended to biosecurity, where the coverage said the field appears ahead of the game. It noted that AI-based tools seem to be causing security threats on a nearly daily basis, partly because society has been slow to recognize potential threats. Google has figured out how to watermark AI-designed proteins, according to the report’s headline. Watermarking is a method of adding identifying information so an item can be traced or distinguished; in this case, the idea is to mark proteins designed with AI tools. The coverage also noted that, as with most things biological, utility goes hand in hand with threats.