Ashish Tandon on Anthropic’s Mythos
Project Glasswing brings together Amazon, Anthropic, Apple, Broadcom, Cisco, CrowdStrike, the Linux Foundation, Microsoft, and Palo Alto Networks.
Bengaluru: Ever since AI firm Anthropic announced Project Glasswing, where select companies get access to Claude Mythos, there has been a growing debate around Artificial Intelligence’s (AI) potential for abuse. Experts have also called for strengthening cybersecurity fundamentals.
Earlier this month, Anthropic said its launch partners will use Mythos Preview for defensive security work and share what they learn so the whole industry can benefit.
Access has also been extended to around 40 additional organisations that build or maintain critical software infrastructure, so they can scan and secure both first-party and open-source systems, it said.
Project Glasswing brings together Amazon, Anthropic, Apple, Broadcom, Cisco, CrowdStrike, the Linux Foundation, Microsoft, and Palo Alto Networks.
Claude Mythos Preview spotted vulnerabilities that have in some cases survived decades of human review and millions of automated security tests, and the exploits it develops are increasingly sophisticated, Anthropic mentioned.
Commenting on Mythos, JP Mishra, Founder & CEO of Deep Algorithms, said it is not a product launch; it is a threshold crossing.
“Nobody is saying this out loud, but it is the first AI that crossed the line where offence permanently and irreversibly beats defence. It has exactly the same power for attackers as it does for defenders. There is no way around that brutal dual-use dilemma. Claude Code is already being weaponised by state groups against several global targets. Mythos amplifies that threat by an order of magnitude.”
Headquartered in Hyderabad, Deep Algorithms is an AI-native cybersecurity company.
He added that a full, sophisticated state-level cyber attack now costs less than $20,000. “Literally anyone can do it. Every power grid, every bank, every piece of old infrastructure on the planet is completely exposed,” he stated.
Anthropic’s choice not to release Mythos is a defining event for the beginning of the era of artificial intelligence, revealing a hard fact that AI’s potential for abuse increases as it advances, Dipesh Ranjan, Senior Vice President at Cyble, said.
According to him, Mythos is a warning that in the race for AI superiority, the creation of advanced capabilities without oversight can be just as dangerous as it can be revolutionary.
Though experts raised various concerns, a few see this as an opportunity. Recently, during Infosys’ Q4 earnings conference, CEO Salil Parekh pointed out both risks and opportunities from Mythos.
Pankit Desai, Co-founder and Ceo, Sequretek, a cybersecurity firm, said Mythos is not an incremental improvement in vulnerability discovery – it is a categorical shift.
“What makes it unlike anything we have seen before is its ability to stitch through multiple attack vectors simultaneously, identifying vulnerability chains that span thousands of permutations across a single application. It found exploitable gaps in OpenBSD, a 27-year-old foundational software trusted in systems as critical as rocket launches. No conventional tool, no VA scanner, no penetration test, no AppSec review, comes close to operating at that scale, complexity, and speed,” he said.
According to him, this is today’s Y2K moment. The potential exposure touches every software application, database, and operating system in use today. Anthropic has, for now, restricted access. But this will not stay contained.
“The moment Mythos goes mainstream, Pandora’s box opens. And there will be no putting the genie back in the bottle,” he added