Prompt Injection is a attack_type tracked across 50 threat clusters and 114 intelligence report mentions on ThreatCluster. First observed October 28, 2025; most recent activity July 17, 2026.
Security firm Tracebit has introduced a novel defense mechanism called 'context bombs' that can thwart AI-powered cyberattacks. By embedding specific strings of text within cloud decoy resources, these context bombs…
Recent reports indicate that hackers are leveraging indirect prompt injection attacks to exploit AI agents, particularly large language models (LLMs). These attacks involve manipulating hidden web content to execute…
Researchers at Pillar Security discovered a critical prompt injection vulnerability in Google's Antigravity IDE, which allows for remote code execution (RCE) by exploiting insufficient input sanitization in the…
A Cisco study has found that 15 major AI models from OpenAI, Anthropic, Google, Amazon, and xAI are significantly more vulnerable to multi-turn prompt injection attacks than previously reported. The research tested…
Recent research highlights significant vulnerabilities in agentic AI systems, which can misclassify adversarial inputs as trusted instructions across six architectural layers. These vulnerabilities stem from trust…
Research from RSAC has revealed that Apple's on-device AI, Apple Intelligence, is vulnerable to prompt injection attacks, allowing attackers to manipulate the model's output. The researchers achieved a 76% success rate…
Recent investigations into AI browsers like Perplexity Comet and ChatGPT Atlas reveal significant security vulnerabilities, particularly prompt injection attacks. These attacks allow malicious actors to embed harmful…
Anthropic reported a 31.5% hijack rate for its Opus 4.8 browser agent before safeguards were activated. This statistic was disclosed in a 244-page system card released on May 28, 2026. The hijack rate indicates that…
On July 15, 2026, OpenAI announced GPT-Red, an automated red-teaming AI model designed to identify vulnerabilities in its language models, particularly against prompt injection attacks. GPT-Red achieved an 84% success…
A vulnerability in Claude Code allows the AI to bypass its deny rules when presented with a chain of over 50 subcommands. Discovered by Adversa, a Tel Aviv-based security firm, this flaw exposes the AI to prompt…