Google wants to compete with Anthropic’s Mythos

Google is making a big push in cybersecurity. At I/O, the company announced that it is inviting a select group of experts to test the CodeMender API, an “AI code protection assistant” that started last October. The difference is that Google is now making the tool more widely available externally – and selling it as a way, as Google DeepMind CTO Koray Kavukcuoglu said, “to help protect the world’s code” by writing and fixing vulnerabilities.

Anthropic Marvel’s announcement of the Claude Mythos Preview looks set to shock the world of AI – and tons more, like top banks and the seat of the Federal Reserve. So, inspired by Anthropic’s news, many companies have been following suit by offering their own powerful AI models designed to uncover unknown security gaps in advanced systems. Anthropic made the Mythos Preview themes too powerful for public release, and supported the company get back in the good graces of the US government after the mention of the risk of the supply chain and the subsequent charge. It also represents the company’s potential for more revenue if things go well with early-stage business users and government agencies. So as the pressure builds for some companies to make a profit ahead of potential IPOs, like OpenAI — or to stay ahead of the growing AI competition, like Google — AI labs are seeing cybersecurity as a revenue driver.

OpenAI quickly followed suit with its own offerings, and now, so does Google. Kavukcuoglu confirmed Seaside in questions that Google has already been discussing with governments and businesses about using CodeMender to evaluate their systems.

Google CEO Sundar Pichai told reporters at a press conference on Monday, “What Mythos has done, I applaud them, is to show that there is value in a very large model in such security cases.