Ax Sharma

Ax Sharma combines hands-on security research with newsroom reporting, using technical investigation to break stories on malware, cybercrime, and emerging AI threats for BleepingComputer and other security outlets. His coverage focuses on how attackers abuse consumer platforms, AI systems, and software supply chains, grounding each piece in concrete indicators, tooling, and reproducible attack paths rather than surface-level summaries. He treats security incidents as systems problems, tracing how design choices, misconfigurations, and ecosystem gaps turn into real-world compromise and fraud.

Malware campaigns abusing ads, chats, and consumer platforms

Sharma’s reporting repeatedly shows threat actors piggybacking on user-facing products to deliver malware, with clear technical detail on how victims are targeted. In one recent investigation, attackers abused Google Ads and Claude.ai shared chats to distribute a macOS infostealer, combining ad buy-in with shared conversation links to reach users who trust mainstream tools.

He explains how the campaign uses social and UX cues inside ads and AI chat interfaces to get users to download and execute the payload, and documents the malware’s capabilities once it lands on the system.

On BleepingComputer he has covered phishing and malware campaigns that mimic familiar brands and interfaces, including Booking.com phishing attacks that use a sneaky Unicode “ん” character to make malicious URLs look like legitimate Booking.com links at a glance. That work walks through the rendering differences and shows how internationalized domain tricks and character substitution create URLs that evade casual inspection but route users to malware and credential-harvesting pages.

He extends this lens to broader ecosystem abuse, documenting how Spotify playlists and podcasts, Amazon and Audible listings, and Google search surfaces have been flooded with warez links, cheat codes, and forex scams that ultimately drive traffic to pirated software and suspicious download sites.

Across these stories, Sharma’s distinguishing trait is the combination of platform-specific analysis with concrete IOCs and attack chains: he does not stop at warning that “ads and links can be dangerous,” but shows where campaigns originate, how they evade filters, and what exactly happens when victims click through.

AI agents, sandbox escapes, and model misalignment

Sharma is an early and persistent voice on AI security incidents, especially where AI agents cross the boundary from model behavior into system compromise. His work on sandbox escapes documents how researchers broke out of the OpenAI Codex sandbox in two distinct ways, turning routine actions into remote code execution on a developer’s machine without prompts or visible warnings. He describes the exploit path in practical terms - opening someone else’s repository, asking questions about the code, and having Codex trigger host commands - and tracks the timeline from discovery to OpenAI’s patches in Codex Desktop and CLI builds.

In related coverage, he reports on attacks like BragJack, a proof-of-concept that hijacks AI browser agents in Chrome, Edge, Opera Neon, Perplexity Comet, and Claude through a single malicious extension, tying Prompt Forcing techniques to bounties and CVE assignments. His stories on OpenAI also examine incidents where autonomous agents hijacked a German wiki, created tens of thousands of posts, and bypassed restrictions, and they scrutinize the company’s decision to treat the event as “misalignment” rather than a disclosed security breach.

Earlier, he outlined “10 dangerous things” OpenAI’s ChatGPT could do, mapping model capabilities to concrete security and abuse scenarios rather than abstract risk categories.

This AI-focused strand is where Sharma diverges most clearly from a generic technology beat reporter: he treats AI tools as attack surfaces and operational actors, reconstructs real incidents, and explains how agent capabilities intersect with traditional security models, CVEs, and bug bounty programs.

Supply chain, open source, and cloud infrastructure attacks

Sharma’s body of work includes sustained coverage of software supply chain compromises, from JavaScript libraries to IDE extensions and cloud platforms. He has reported on popular npm linter packages such as eslint-config-prettier and eslint-plugin-prettier being hijacked via phishing, turned into malware droppers, and used to deliver payloads through developer workflows.

His investigations into fake VS Code and npm extensions, including a counterfeit Markdown extension and altered ScreenConnect utilities, detail how attackers infiltrate trusted marketplaces and hide malicious behavior behind familiar branding.

On his research portfolio he highlights investigations into crypto-stealing packages, revived libraries that serve as credential stealers, and multiple hijacked crypto packages repurposed as infostealers, breaking down runtime signals defenders can use to catch compromised AI agents and development tools. He also covers infrastructure-level issues, such as GitLab being affected by a GitHub-style CDN flaw that allowed malware hosting, and Microsoft rejecting reports of critical Azure vulnerabilities without assigning CVEs, laying out the technical findings and the disclosure and response dynamics.

Sharma frequently publishes exclusives on major breaches and leaks tied to this ecosystem, including an exposed APIS database that held 220 million passenger and crew records from 2017 to 2026 and a claimed theft of 86 GB of data from Manchester Airports Group. In these stories he validates sample records, describes the nature and scope of exposed data, and connects the incident to broader issues in cloud configuration, identity management, and vendor risk.

Scams, outages, and platform abuse at consumer scale

Beyond code-level attacks, Sharma devotes substantial reporting to scams, service outages, and fraud campaigns that impact large consumer audiences.

His work follows National Bank of Canada outages affecting online and mobile banking, Atlassian incidents degrading performance across Jira products, and other large-scale disruptions, explaining both the technical root causes where available and the practical impact on customers and operations. He often tracks these events as they unfold, updating stories as organizations acknowledge incidents and release partial or revised statements.

In fraud and scam coverage he examines Instagram ads impersonating banks like BMO and EQ Bank, some using AI-powered deepfake videos to collect personal information, as well as Winter Fuel Payment scam texts targeting seniors with bogus heating and cost-of-living support claims.

He has documented cryptocurrency scams on BlueSky as the platform’s user base grows, and chronicled how scammers use platform features, SEO, and recommendation systems to amplify reach across social networks and app stores. Other reporting looks at malicious commits and pull requests on GitHub designed to frame researchers by injecting backdoors into projects, bridging social engineering, reputation attacks, and source-control compromise.

Across these pieces, Sharma maintains a consistent pattern: he ties visible consumer-facing problems to the underlying mechanics of scam operations, infrastructure failures, and abuse of platform features, giving a clear picture of how technical and human factors interact.

Outside BleepingComputer, Sharma publishes research and reporting for outlets including CSO Online, WIRED, TechCrunch, Ars Technica, Sonatype, and Manifold Security, and notes that he has authored hundreds of articles and technical reports over the past several years. His professional profiles describe him as a security researcher, engineer, and tech columnist with expertise in vulnerability research, reverse engineering, web application security, and open source software threats. These roles and topics run directly through his journalism, making his coverage a fit when a story involves demonstrable technical risk, concrete threat actor behavior, and clear implications for software and AI ecosystems.