Researchers at Tenet Security have disclosed Agentjacking, a new attack that turns AI coding assistants like Claude Code, Cursor, and Codex into tools for running an attacker's code on a developer's machine. The trick abuses Sentry, a widely used error-tracking service: anyone can submit a fake error event using a project's DSN, a public write-only key embedded in website code, and the AI agent, fetching that event through Sentry's MCP integration, cannot tell the malicious instructions from real diagnostics and runs them with the developer's privileges. No phishing, malware, or server breach is needed, and it bypasses traditional controls because every step is technically authorized. Tenet found 2,388 exposed organizations.
Researcher RyotaK has disclosed a now-patched flaw in Anthropic's Claude Code GitHub Action, which drops Claude into CI/CD to triage issues and review PRs with broad repo permissions. The action's trigger check waved through any actor whose name ended in [bot] - but anyone can register a GitHub App and use its token to open an issue on a public repo. Agent mode lacked the human-actor check tag mode had. The attacker then used indirect prompt injection in an issue to make Claude read /proc/self/environ and write back the OIDC credentials, which can be replayed for an installation token with write access. Anthropic's example workflow shipped with allowed_non_write_users: '*'.
SafeBreach's Or Yair has demonstrated Fake Context Alignment, a technique that hijacks Google Gemini's voice assistant on Android through malicious notifications from apps like WhatsApp and Slack - no malicious app on the phone required. Gemini's Utilities feature reads and acts on notification text as if it were instructions, an attack surface Yair calls 'effectively infinite.' The bypass runs two illusions at once: it poses the real authorization question in a language the victim does not speak, defeating Google's post-Invitation prompt-injection mitigations. It can fake a boss's message, open windows, force a Zoom call, or poison long-term memory. Google has patched it; no CVE was assigned.
Permiso Security has disclosed ChatGPhish, a vulnerability in OpenAI ChatGPT that abuses the assistant's implicit trust in Markdown links and images sourced from third-party pages it has just summarized. The chatgpt.com response renderer auto-fetches those images and surfaces the links as live clickable elements inside the trusted assistant UI. An attacker who appends a small payload to any web page a victim later asks ChatGPT to summarize can leak the victim's IP, User-Agent, and Referer via attacker-hosted images, render fake system-style security alerts, plant malicious clickable links, and serve a QR code from an S3 bucket to bypass desktop URL filters via the victim's phone.
Two related stories show AI-powered developer tools becoming a fresh attack surface. First, Pillar Security disclosed a now-patched vulnerability in Google's agentic IDE Antigravity that allowed prompt injection to escape the Strict Mode sandbox and achieve arbitrary code execution. The flaw combined Antigravity's file-creation capability with insufficient input sanitization in its find_by_name tool: injecting the -X (exec-batch) flag via the Pattern parameter forced the underlying fd utility to execute arbitrary binaries against workspace files. An attacker could stage a malicious script then trigger it through a seemingly legitimate search - no user interaction needed once the prompt injection lands. The attack can be delivered via indirect prompt injection: a user pulls a harmless-looking file from an untrusted source containing hidden comments that instruct the AI agent to stage and trigger the exploit. Google patched on February 28. Second, Manifold Security researchers showed a Claude-powered GitHub Actions workflow (claude-code-action) can be tricked into approving and merging malicious pull requests by setting Git's user.name and user.email to match a trusted developer (in the demo: Andrej Karpathy). On first submission Claude flagged for manual review. On resubmission, Claude approved it - the AI overrode its own earlier judgment on retry. The common thread: AI agents cannot treat attacker-controllable metadata as a trust signal, and non-determinism across retries means you cannot build a security control on an AI that changes its mind.
TrustStrike Labs