Microsoft (MSFT) is turning to AI to help protect users from attacks and hacks, leveraging a new class of models that can find and exploit software flaws faster than any person could on their own.
Part of that includes the company’s new Microsoft Security multimodal agentic scanning harness, or MDASH. The platform, which Microsoft said is in an expanded preview, uses AI agents to scan for software vulnerabilities that hackers can exploit.
Speed is a key component of cybersecurity. The faster a company can detect a software flaw, the sooner it can fix the flaw before hackers can exploit it.
“AI brings a lot of superpowers to defenders, but also AI in the hands of cyber attackers is a very powerful tool,” Vasu Jakkal, corporate vice president of security at Microsoft, told Yahoo Finance.
“The time between vulnerability discovery to exploitation essentially collapses … and if we don’t build on this technology with context for defenders, that is going to continue to be a very asymmetric cyber war across the industry,” she added.
MDASH uses a collection of more than 100 configurable AI models to search for exploitable bugs, which the company said ensures it’s not too dependent on a single model and provides model variety between state-of-the-art options to more cost-efficient offerings.
While generative AI certainly raised cybersecurity concerns over the past few years, Anthropic’s (ANTH.PVT) Claude Mythos Preview has set off alarm bells across the industry.
Microsoft’s new Microsoft Security multimodal agentic scanning harness, or MDASH, uses AI agents to scan for software vulnerabilities that hackers can exploit. (Beata Zawrzel/NurPhoto via Getty Images) · NurPhoto via Getty Images
The AI model, which Anthropic revealed in April, is surprisingly adept at finding flaws in existing software, including all major web browsers and operating systems, despite not being developed for cybersecurity purposes.
To prevent widespread security issues, Anthropic initially partnered with a group of roughly 40 companies, including Microsoft, to provide them with access to Mythos so they could use it to find and patch holes in their software.
On Tuesday, Anthropic announced it is expanding access to Mythos to some 140 additional companies across multiple countries.
Like Anthropic, OpenAI (OPAI.PVT) has its own cybersecurity-centric AI model, called GPT-5.4 Cyber.
Sign up for Yahoo Finance’s Week in Tech newsletter. · Yahoo Finance
Email Daniel Howley at dhowley@yahoofinance.com. Follow him on Twitter at @DanielHowley.
Read the latest financial and business news from Yahoo Finance