Skip to content
Technology & InnovationLatest

USV Bets on AI Agents for an Antifragile Internet Amidst Rising Vulnerabilities

Published

The internet's increasing vulnerability, starkly revealed by recent attacks, is now met with a strategic bet by USV that AI agents, paradoxically, hold the key to building a more antifragile digital future.

MissedBlock Desk · 4 min read

USV Bets on AI Agents for an Antifragile Internet Amidst Rising Vulnerabilities

The internet’s increasing vulnerability, starkly revealed by recent attacks, is now met with a strategic bet by USV that AI agents, paradoxically, hold the key to building a more antifragile digital future. In October 2026, a series of incidents, including a PHP vulnerability affecting an estimated 500 million hosts and a $6.3 billion theft from Binance’s hot wallet, underscored the precarious state of online infrastructure. USV, an investment firm, views AI agents as a critical factor in this evolving landscape, seeing them as both a potent threat and a potential solution for hardening the internet.

According to USV, AI agents are fundamentally altering the internet’s structure in three key ways: as weapons, as targets, and as impersonators. As weapons, AI agents are dramatically lowering the cost and complexity of exploiting existing vulnerabilities. For instance, a previously hidden PHP vulnerability was weaponized within hours using advanced AI models, a task that would have previously taken significantly longer. This trend pushes the cost of vulnerability exploitation towards zero, posing a substantial risk to legacy systems.

Simultaneously, AI agents and the code they generate are becoming targets themselves. Attacks like prompt injection, which manipulates AI models into unintended actions, are becoming more prevalent. Furthermore, AI agents are increasingly impersonating human traffic. Cloudflare data indicates that by 2026, half of all internet traffic was not human, with CEO Matthew Prince predicting automated traffic to eventually be 1000 times that of human traffic. This surge in non-human activity, exemplified by millions of malicious OpenAI agent requests targeting Wikimedia Foundation websites in October, necessitates new paradigms for traffic management and security.

USV’s Antifragility Strategy

In response to these challenges, USV proposes a strategic shift towards building an antifragile internet—systems that improve and adapt when faced with disruption. USV’s approach focuses on three countermeasures: observation, execution, and prevention. While much of the current capital and company focus is on observation (monitoring) and execution (responding to incidents), USV emphasizes the critical need for prevention. This involves supporting methods that eliminate entire classes of bugs, leveraging the network effect of agent security logs, and integrating security directly into tools and workflows.

The Evolving Ecosystem of AI Security

The market is already responding to this evolving threat landscape. A new stack of security tools is emerging around AI agents, including code review platforms like Greptile, scanning tools such as Semgrep, dependency checkers like Socket, and incident response services like Resolve AI. New deployment layers, such as agent sandboxes and authorization tools, are also forming, aiming to create an antifragile developer stack.

AI agents are also accelerating code migration. Previously, manually rewriting large legacy codebases into more secure languages like Rust was an arduous and expensive undertaking. However, AI tools like GitHub Copilot have demonstrated the ability to migrate hundreds of thousands of lines of code in a matter of months with minimal human intervention. This acceleration could trigger a significant ‘code migration wave’ as organizations seek to move away from vulnerable legacy systems.

Furthermore, privacy and security are increasingly being embedded as features within consumer applications rather than offered as standalone products. Companies like Privy and Turnkey are abstracting complex cryptographic operations into user-friendly interfaces, making advanced security features accessible. Technologies such as zero-knowledge proofs and Trusted Execution Environments (TEEs) are being integrated into familiar applications, including messaging services and digital identity solutions.

Navigating Future Uncertainties

Despite these advancements, significant uncertainties remain. The reliability of AI-generated code migrations and the potential for new bugs are areas of concern. The effectiveness of formal verification systems, while promising, depends heavily on the quality of specifications. The long-term implications of portable agent reputation and the potential for centralized control in agent-to-agent communication protocols also present open questions. USV’s investment thesis is built on navigating these complexities, betting that the proactive, AI-driven development of antifragile systems will ultimately define the future of internet security.

Why This Matters

The materials describe a narrow update: USV, an investment firm, is betting on AI agents to make the internet more antifragile, moving away from its current state of vulnerability exposed by recent attacks. The exact timeline for the predicted ‘code migration wave’.

Broader Context

Source materials place the market analysis in this context: The article discusses the increasing vulnerability of the internet, citing specific attacks in October 2026 involving PHP vulnerabilities and a major cryptocurrency exchange hack.

USV Bets on AI Agents for an Antifragile Internet Amidst Rising Vulnerabilities · MissedBlock