History shows us that whenever technology increases the speed of the attack, the only way to survive is to increase the speed of the adaptation | Photo Credit: MF3d

In May 1940, the world’s military elite believed that the Belgian fortress of Eben-Emael was the most impregnable structure ever built. It was a masterpiece of concrete and steel, designed to withstand any imaginable siege. Yet, it fell in just 15 minutes. It wasn’t destroyed by a massive army, but by a handful of paratroopers using a new, secret technology: the shaped charge. The defenders were prepared for a war of attrition; they were erased by a war of speed and specific intelligence.

Last month, the global cybersecurity landscape hit its own Eben-Emael moment. For 27 years, a subtle flaw lay dormant in the code of OpenBSD — an operating system whose entire reputation is built on being the world’s most secure foundation. For nearly three decades, the finest human auditors, state-sponsored “red teams”, and automated scanners looked at that code and saw a fortress. Then came Mythos.

Developed by Anthropic, Mythos is not a chatbot; it is a Reasoning Agent for Cyber-Offense. It did not just “suggest” a bug; it understood the systemic logic of the flaw and autonomously executed a 32-step exploit in seconds. It achieved an 83.1 per cent success rate in reproducing known hacks on its first attempt. This is what I call the “End of Surprise”. When a machine can hunt for “zero-days” (previously unknown flaws) at a scale and speed that humans cannot register, the concept of a periodic security audit is no longer a safety measure. It is a dangerous hallucination.

The shockwaves from Mythos have fundamentally rewritten the geopolitical script. After months of advocating for a “light touch” approach to AI, the Trump administration has executed a stunning pivot towards what can only be described as the Nationalisation of Intelligence.

Reports from Washington suggest a new framework where the Pentagon and the NSA will act as the ultimate gatekeepers for “frontier” AI models. By mandating military-led safety tests before any public release, the US is treating high-level AI as a dual-use weapon system, akin to nuclear enrichment technology. This is “Fortress America” in digital form. The realisation has dawned: in a world of Mythos-class agents, a leaked model is not a commercial loss; it is a national security catastrophe.

For India, however, following this defensive, gatekeeping model would be a historic strategic blunder. We are a nation built on Digital Public Infrastructure (DPI) — from the UPI rails that power our economy to the Aadhaar stack that defines our identity. Our strength is our openness. If we wait for Western “safety certificates” while our systems remain exposed to the tools they have already unleashed, we are essentially building our own Maginot Line and waiting for the Blitzkrieg.

The Indian valuation cliff

The threat to India is not just a matter of national security; it is a matter of national business. Our IT services giants — TCS, Infosys, Wipro — have spent three decades building a global empire on the “billable hour”. Thousands of engineers are employed to conduct manual audits, patch vulnerabilities, and maintain legacy code.

Mythos represents a “Valuation Cliff” for this model. If an AI can perform a month’s worth of security auditing in five seconds at near-zero marginal cost, the human-led audit becomes a bottleneck rather than a service.

The Indian IT sector must pivot immediately from being “service providers” to “resilience architects”. We must move from charging for the time it takes to fix a bug to charging for the integrity of a system that can defend itself.

The physicist’s lens

As a physicist, I view this crisis through the lens of Epistemic Drift. This is the point at which the complexity of our digital foundations exceeds our human ability to verify their safety. We are building structures we can no longer see through.

When we can no longer trust the very foundation of our software because an AI has found a 27-year-old “hole”, the social contract of the digital age begins to fray.

In physics, entropy is the natural slide into disorder. In the digital world, Mythos is an entropy engine. It can find and weaponise disorder faster than we can organise a defence. To counter this, we must stop thinking about “safety” as a lock on a door and start thinking about “resilience” as a biological immune system.

I have often spoken of “sovereign resilience”, but in the wake of Mythos, this must evolve from a policy suggestion into a biological necessity for the state. This requires a fundamental shift in our national architecture.

First, we must acknowledge that sovereignty is compute. India cannot be a “Tier-2” participant in restricted US programs like Project Glasswing. We must use initiatives like BharatGen to build sovereign “Defensive LLMs”. These are not chatbots for the public; they are internal “Guard Dogs” trained specifically on the India Stack to hunt for vulnerabilities and simulate attacks 24/7. We must be the first to find our own flaws.

Second, we must move toward Autonomous Self-Healing. In a world of autonomous attacks, a human “admin” logging in to fix a server is a relic of the past. Our power grids, our telecommunications, and our financial stacks must be redesigned to detect an anomaly and “re-wire” their own logic in real-time. This is not about building better firewalls; it is about building a system that can “bleed” and still keep running.

Finally, we need a Post-Audit Regulatory Framework. Our regulators, including the CS Setty Committee looking into banking risks, must stop focusing on the “prevention” of attacks. In the age of Mythos, the attack is an automated certainty. The only metric that matters now is Time-to-Resilience. How many milliseconds does it take for a bank to detect a breach, isolate it, and return to a “verified” state? That is the only audit that matters in 2026.

The new atmosphere

History shows us that whenever technology increases the speed of the attack, the only way to survive is to increase the speed of the adaptation. The tank made the trench obsolete; the aircraft made the battleship a target; and now, autonomous AI has made the static firewall a ghost of the past.

The Mythos event is not a software update; it is a change in the digital atmosphere. It is the end of the era where we could afford to be surprised. For India to lead, we must stop trying to build better walls. We must build a more agile, self-correcting nation.

In the age of autonomous intelligence, the only true fortress is the ability to adapt faster than the machine. The old walls have already been bypassed. It is time for India to learn how to move.

The writer is a physicist at the University of North Carolina at Chapel Hill, and a columnist on AI, infrastructure, and global systems

Published on June 11, 2026