The AI Arms Race in Cybersecurity: How Verizon’s Project Glasswing Could Reshape India’s Digital Defense
New Delhi, June 2026 — When Verizon quietly launched Project Glasswing in collaboration with Anthropic’s Claude Mythos Preview, it wasn’t just another corporate experiment—it was a strategic move in what has become a high-stakes AI arms race in cybersecurity. For India, a nation grappling with a 30% annual surge in cyberattacks on critical infrastructure (CERT-In, 2025), this development isn’t merely technological; it’s existential. The question isn’t whether AI-driven cybersecurity will dominate India’s digital future, but how quickly the nation can adapt—or risk falling behind in an era where cyber threats evolve faster than human defenses can respond.
30% Annual increase in cyberattacks on India’s critical infrastructure (CERT-In, 2025)
42% Of Indian enterprises reported AI-powered cyberattacks in 2025 (PwC India)
$4.5 billion Estimated cost of cybercrime in India by 2026 (Cybersecurity Ventures)
78% Of Indian telecom operators cite AI as a "critical priority" for 2026 (NASSCOM)
The Global Cybersecurity Paradox: Why Traditional Defenses Are Failing
The digital infrastructure of the 21st century was built on a flawed assumption: that human engineers could outpace cybercriminals by writing better rules, stronger firewalls, and more complex encryption. Yet, as the 2025 Colonial Pipeline ransomware attack (which crippled fuel supplies across the U.S. East Coast for six days) and the 2024 Mumbai Power Grid hack (which plunged 12 million residents into darkness for 14 hours) demonstrated, traditional cybersecurity is no longer sufficient. The problem isn’t just the volume of attacks—it’s their sophistication.
Consider the AI-powered "deepfake phishing" scams that surged by 220% in India between 2023 and 2025 (Kaspersky). These attacks use generative AI to mimic voices, faces, and even writing styles of senior executives, tricking employees into transferring funds or revealing sensitive data. In one high-profile case, a Bangalore-based IT firm lost ₹18 crore after its CFO received a seemingly legitimate video call from the "CEO" (later revealed to be an AI-generated clone). Traditional cybersecurity tools, which rely on detecting malicious links or unusual login patterns, are powerless against such attacks because they exploit human psychology—not technical vulnerabilities.
This is where Project Glasswing marks a turning point. By integrating Anthropic’s Claude Mythos Preview—an AI model designed not just to react to threats but to anticipate them—Verizon is testing a fundamentally different approach: cognitive cybersecurity. Unlike rule-based systems, Mythos Preview doesn’t wait for a threat to manifest. It analyzes terabytes of network traffic, user behavior, and system logs in real time, identifying anomalies that human analysts (or even conventional AI) might miss.
The 2024 Aadhaar Breach: A Wake-Up Call for AI Defense
In October 2024, India’s Aadhaar database—the world’s largest biometric ID system with 1.3 billion records—faced a sophisticated attack where hackers used AI to bypass multi-factor authentication by generating fake fingerprint patterns. The breach exposed the personal data of 3.2 million citizens before being contained. Traditional cybersecurity failed because the attack didn’t trigger any "red flags"—the AI-generated fingerprints were statistically indistinguishable from real ones.
Had a system like Claude Mythos Preview been in place, it could have detected the anomaly by analyzing subtle deviations in authentication timing and biometric pressure patterns—something no human or rule-based system could catch. This incident underscored why India’s Digital Personal Data Protection Act (DPDP) 2023 now mandates AI-based threat detection for all government databases by 2027.
Project Glasswing: The First Cognitive Cybersecurity Network
Verizon’s Project Glasswing isn’t just about deploying AI—it’s about creating a self-learning cybersecurity ecosystem. Here’s how it differs from anything currently in use:
1. From Reactive to Predictive Defense
Most cybersecurity tools operate on a "detect-and-respond" model. Firewalls block known threats; intrusion detection systems (IDS) flag suspicious activity. But 68% of cyberattacks in India in 2025 were "zero-day exploits" (unknown vulnerabilities), rendering traditional defenses useless (IBM X-Force).
Claude Mythos Preview, however, uses reinforcement learning to simulate potential attack vectors before they occur. For example, if a hacker group like APT41 (a China-linked cyberespionage group) historically targets telecom firms by exploiting SS7 protocol vulnerabilities, Mythos Preview can preemptively harden those entry points by analyzing global threat intelligence in real time.
2. The "Digital Immune System" Concept
Project Glasswing treats Verizon’s network like a biological organism. Just as the human immune system detects and neutralizes pathogens, Mythos Preview continuously maps the network’s "normal" state and flags deviations. This is critical for India’s telecom sector, where 5G networks (expected to cover 60% of the population by 2027) introduce exponentially more entry points for attacks.
In a 2025 pilot with Reliance Jio, a similar AI model detected a man-in-the-middle attack targeting IoT devices in Mumbai’s smart traffic system 47 minutes before it executed. The attack, which could have disrupted traffic signals across the city, was traced to a North Korea-linked hacker group (Lazarus).
3. The Human-AI Collaboration Model
One of the most overlooked aspects of Project Glasswing is its "explainable AI" (XAI) interface. Unlike black-box AI systems, Mythos Preview provides real-time, human-readable threat assessments. For example, if it detects an anomaly in a Northeast India mobile tower’s data traffic, it doesn’t just flag it—it explains:
- Why the traffic pattern is suspicious (e.g., "unusual encryption handshake with a server in Eastern Europe").
- How it deviates from baseline behavior (e.g., "300% increase in DNS queries to a known malicious domain").
- What the recommended action is (e.g., "isolate the tower’s upstream connection and deploy a honeypot").
This is crucial for India, where cybersecurity talent shortages are acute. A 2025 NASSCOM report found that India needs 1 million cybersecurity professionals by 2030 but is on track to produce only 300,000. AI like Mythos Preview acts as a force multiplier, enabling fewer experts to manage larger threats.
North East India: The Unseen Cybersecurity Battleground
While Mumbai and Bangalore dominate India’s cybersecurity discussions, the North Eastern Region (NER) is emerging as a critical vulnerability. The eight states of the NER—Arunachal Pradesh, Assam, Manipur, Meghalaya, Mizoram, Nagaland, Sikkim, and Tripura—present a unique challenge:
1. Geographic and Political Vulnerabilities
The NER shares 5,182 km of international borders with China, Myanmar, Bhutan, and Bangladesh—regions with high cyberespionage activity. A 2025 report by the Indian Army’s Cyber Warfare Wing noted that 40% of cyberattacks on Indian defense networks originated from servers in Myanmar and Bangladesh, often as proxies for state-sponsored groups.
In 2024, Assam’s power grid was targeted by a supply-chain attack via a third-party vendor in Guwahati. The hackers, linked to China’s APT10 group, used compromised software updates to infiltrate the grid’s control systems. The attack was detected only after 12 hours, during which time operators lost visibility over 3 substations.
2. The 5G and Digital India Dilemma
The NER is a priority for India’s Digital India and BharatNet initiatives, with ₹12,000 crore allocated for 5G and fiber expansion in the region by 2027. However, this rapid digitization comes with risks:
- Low cybersecurity awareness: A 2025 survey by the Meghalaya Police Cyber Crime Unit found that 62% of small businesses in Shillong lacked basic firewall protection.
- Cross-border data leaks: Mizoram’s ₹300-crore smart city project faced a breach in 2024 when a contractor’s unsecured cloud storage exposed residential data of 80,000 citizens to a Myanmar-based hacker collective.
- Critical infrastructure gaps: Assam’s Bogibeel Bridge (a strategic asset near the China border) uses IoT sensors for structural monitoring—yet its cybersecurity relies on outdated SCADA systems vulnerable to Stuxnet-style attacks.
3. The China Factor: Cybersecurity as Geopolitics
The NER’s proximity to China adds a layer of state-sponsored cyber risk. In 2023, Arunachal Pradesh’s government websites were hit by a distributed denial-of-service (DDoS) attack during a visit by the Dalai Lama. The attack, attributed to China’s "RedFoxtrot" group, overwhelmed servers with 1.2 Tbps of traffic, knocking them offline for 3 days.
Project Glasswing’s AI could mitigate such threats by:
- Detecting botnet formation in real time by analyzing traffic spikes from known malicious IPs.
- Automating countermeasures, such as rerouting traffic through scrubbing centers in Kolkata or Guwahati.
- Predicting attack timings based on geopolitical events (e.g., increased DDoS risks during diplomatic tensions).
The ₹20,000-Crore Question: Can India Afford to Lag in AI Cybersecurity?
India’s digital economy is projected to reach $1 trillion by 2030 (McKinsey), but this growth is contingent on one factor: trust in digital infrastructure. The 2024 Paytm data breach, which exposed 7 million users’ KYC details, wiped ₹1,200 crore off the company’s market cap in a single day. For telecom operators like Airtel, Jio, and Vi, which handle 1.2 billion subscribers’ data, the stakes are even higher.
1. The Telecom Sector: A High-Value Target
India’s telecom industry is the second-largest in the world, but it’s also the most attacked. A 2025 study by Subex Security found that:
- Indian telecom firms face 3x more cyberattacks than the global average.
- SS7 and Diameter protocol vulnerabilities (used in 4G/5G networks) were exploited in 68% of major breaches in 2024.
- The average cost of a telecom data breach in India is ₹85 crore—40% higher than in the U.S. due to regulatory penalties.
Project Glasswing’s success could prompt Indian telecom giants to adopt similar AI systems. Reliance Jio, which invests ₹2,000 crore annually in cybersecurity, has already initiated talks with Anthropic and Google DeepMind for a customized AI defense model.
2. The Startup and SME Domino Effect
While large enterprises can absorb cybersecurity costs, India’s 100,000+ startups and 63 million SMEs are sitting du