Skip to main content
Mallory
← Back to News
ai-platform-securityidentity-impersonation-fraudcredential-access-methodpackage-repository-poisoning

Zero-Click Prompt Injection Hijacked ChatGPT Atlas and Claude Browser Agents

Emerging Technology Risk · 7 sources · Updated 3mo ago · First seen Feb 17, 2026

Zero-Click Prompt Injection Hijacked ChatGPT Atlas and Claude Browser Agents

Zenity disclosed zero-click attacks that hijacked OpenAI’s ChatGPT Atlas and Anthropic’s Claude in Chrome by embedding malicious instructions in content the agents were allowed to read, including X posts and email messages. In the Atlas demonstrations, a planted X comment exploited intent collision to redirect a benign task into phishing messages sent from the victim’s authenticated WhatsApp Web session, and a separate chain steered Atlas into making an unauthorized Amazon purchase by abusing Amazon’s Rufus assistant to finish the order. Researchers said Atlas’s prompt-based guardrails and classifiers could be bypassed by distributing instructions across page elements, using alternate language text, and making the malicious action appear aligned with the user’s original goal.

A parallel attack against Claude in Chrome used indirect prompt injection in an email and a rogue NPM package served from a malicious CDN to escalate into JavaScript execution inside the victim’s active browser sessions. Zenity said the chain enabled Gmail inbox theft, persistent Google Drive sharing abuse, and takeover of Slack, X, and Claude.ai accounts by intercepting verification codes from Gmail. The findings were framed as an architectural risk in agentic browsers that can act across authenticated sessions, effectively recreating CSRF-like cross-origin abuse at the AI agent layer; Zenity said it reported the Atlas issues to OpenAI and the Claude issues to Anthropic months earlier, but the attacks remained exploitable or were treated as informational at publication time.

New surface, new adversariesTrack how attackers are adapting to this technology.
Track this risk
Event timeline

How this story unfolded

Events from the most recent confirmed update back to the earliest known activity.

10 events · 2008 – Feb 17, 2026
  1. Reported Aug 6, 20263mo ago

    SecurityWeek reports Zenity's AI browser hijacking findings

    SecurityWeek reported Zenity's disclosure of zero-click attacks against ChatGPT Atlas and Claude in Chrome, highlighting cross-session account takeover, phishing, and unauthorized purchases, and noting the issues remained unpatched or classified as informative at the time of reporting.

    Zero-Click AI Browser Hacking: Claude and ChatGPT Atlas Hijacked via Emails, X Posts - SecurityWeek
  2. Reported Aug 5, 20264mo ago

    Zenity publishes Grand Theft Atlas research

    Zenity published research on two zero-click Atlas hijacking techniques based on intent collision, showing phishing via WhatsApp Web and unauthorized Amazon purchases completed with Amazon Rufus.

    Grand Theft Atlas: How We Hijacked ChatGPT's AI Browser | Zenity Labs
  3. Reported Aug 5, 20264mo ago

    Zenity publishes Claude in Chrome takeover research

    Zenity published a detailed write-up describing how indirect prompt injection in a malicious email could lead Claude's Chrome extension to execute attacker-controlled JavaScript, exfiltrate Gmail, abuse Google Drive sharing, and take over Slack, X, and Claude.ai accounts.

    Security Research | Claude in Chrome: From alert(1) to Full Account Takeover | Zenity Labs
  4. Feb 17, 20269mo ago

    OpenAI acknowledges Atlas report

    OpenAI acknowledged Zenity's Atlas disclosure and recognized meaningful risks from prompt injection in agentic environments.

    Grand Theft Atlas: How We Hijacked ChatGPT's AI Browser | Zenity Labs
  5. Jan 27, 202610mo ago

    Anthropic closes initial Claude report as informative

    Anthropic closed Zenity's initial Claude in Chrome report as informative rather than remediating it, according to Zenity's write-up.

    Security Research | Claude in Chrome: From alert(1) to Full Account Takeover | Zenity Labs
  6. Jan 13, 202610mo ago

    Anthropic closes duplicate Claude report

    Anthropic closed one of Zenity's Claude in Chrome reports as a duplicate and said the issue was ineligible for its vulnerability disclosure program.

    Security Research | Claude in Chrome: From alert(1) to Full Account Takeover | Zenity Labs
  7. Jan 12, 202610mo ago

    Zenity submits second Claude report to Anthropic

    Zenity submitted a second report to Anthropic on January 12, 2026 covering Claude in Chrome exploitation details; SecurityWeek also summarizes that Claude findings were reported in January 2026.

    Security Research | Claude in Chrome: From alert(1) to Full Account Takeover | Zenity Labs
  8. Jan 11, 202610mo ago

    Zenity reports Atlas attacks to OpenAI

    Zenity disclosed its ChatGPT Atlas findings to OpenAI, including zero-click hijacking scenarios involving WhatsApp phishing and unauthorized Amazon purchases.

    Grand Theft Atlas: How We Hijacked ChatGPT's AI Browser | Zenity Labs
  9. Dec 27, 202511mo ago

    Zenity reports Claude in Chrome issues to Anthropic

    Zenity reported security issues in Anthropic's Claude Chrome extension through HackerOne on December 27, 2025, describing prompt-injection-driven abuse of the extension's browser capabilities.

    Security Research | Claude in Chrome: From alert(1) to Full Account Takeover | Zenity Labs
  10. 2008

    uTorrent CSRF exploit cited as real-world example

    The OWASP reference cites a 2008 uTorrent exploit as a real-world CSRF attack that used forged GET requests to trigger malware downloads at scale.

    Cross Site Request Forgery (CSRF) | OWASP Foundation
Linked entities

Related entities

Vulnerabilities, threat actors, malware, products, organizations, breaches, and observables Mallory has linked to this story. Indicator values are masked here and available in full in the app.

Open in app →

Affected products

1 linked
Claude

Organizations

2 linked
AnthropicOpenai

Observables

1 linked

Indicator values are masked on this page. See the values in Mallory Domains, IPs, hashes, and URLs are exportable to your SIEM.

Do you run this?Follow how adversaries are adapting to this technology, and where it touches your stack today.
Track This Risk
Mallory platform

See the full picture, correlated to your attack surface.

This page covers what’s public. Mallory adds the parts that aren’t: which of your assets are affected, which threat actors are using it right now, and what to deploy next.

Exposure mappingMap indicators from this story to your assets and identify affected systems in minutes.
Threat actor evidenceEvery observed campaign, victim, and pivot linked to actors named in this story.
Associated malwareMalware, exploits, and IOCs connected to the activity described here.
Detection signaturesYARA, Sigma, and Snort rules deployed to your SIEM as soon as they’re published.
Scheduled alertsGet matching new stories delivered to your team as they break, not the next morning.
AI threadsAsk questions about this story and take action on the answers.