Google launches AI Threat Defense to take on Anthropic Mythos and OpenAI Daybreak

Google has introduced AI Threat Defense, an AI-powered cybersecurity platform that prioritises real-world threats and automated fixes. The platform aims to cut alert overload by prioritising exploitable flaws and verified fixes.

In Short

• Google unveils AI-powered cybersecurity tool for businesses and developers
• Tool prioritises real threats instead of flooding teams with alerts
• Launch comes after Anthropic Mythos and OpenAI Daybreak rollouts

Amid growing concerns around powerful AI tools that could potentially hack any software or operating system, Google has introduced its own AI-powered cybersecurity tool. The new platform is called Google AI Threat Defense. Google’s new tool arrives after Anthropic launched its cybersecurity-focused AI model Claude Mythos in April and OpenAI introduced the Daybreak platform, built on GPT-5.5, earlier in May 2026.

While Claude Mythos has raised concerns among companies and governments around the world, including in India due to its capabilities to discover huge numbers of vulnerabilities overnight, Google describes its new tool as an AI-powered cybersecurity solution that continuously monitors and stops AI-driven threats before they can affect businesses.

“AI Threat Defense helps organisations actively predict attack paths, prioritize the most significant threats, and deploy verified fixes faster than adversaries can exploit them,” Francis deSouza, COO of Google Cloud and President of Security Products, wrote in a blog post.

But while both Anthropic and OpenAI largely focused on raw model intelligence and large-scale code scanning, Google is taking a different approach.

Google says cybersecurity teams are drowning in AI alerts

According to Google, one of the biggest problems with current AI cybersecurity systems is that they overwhelm security teams with thousands of AI-generated alerts. Anthropic Mythos and OpenAI Daybreak can rapidly discover huge numbers of vulnerabilities overnight. However, not every vulnerability is actually dangerous in the real world.

Google AI Threat Defense is trying to solve that problem.

The platform integrates code scanning with cloud-security platform Wiz and checks whether a vulnerability is actually reachable from the internet or exposed through live network configurations.

If a high-severity flaw exists in the code but is completely isolated, the system lowers its priority. This allows developers and security teams to focus only on active threat that attackers could realistically exploit.

“Unlike other model providers that simply hand security teams a massive, unprioritised list of AI-generated alerts, we deliver prioritised fixes to accelerate remediation and secure the Defender’s Advantage,” Francis deSouza wrote in the blog post.

Google is also using AI agents to rewrite old code

Google AI Threat Defense continuously performs deep-dive scans across millions of lines of corporate code. The system uses lighter and more cost-effective AI models for broad continuous monitoring, while Gemini frontier models are deployed for high-risk assets and complex exploitability analysis.

Another major difference is how Google handles fixes.

While OpenAI’s system mainly proposes patches, Google says it uses autonomous AI agents under human supervision to actively rewrite legacy code into modern and memory-safe programming languages.

The platform can also analyse dependencies and automatically generate tests to verify whether a fix actually works before deployment.

- Ends