Prompt Injection Risks Expand From LLMs to Embodied AI via Environmental Text
Researchers and security commentators warned that prompt injection remains a fundamental weakness in today’s large language models (LLMs), where carefully crafted inputs can override guardrails and elicit restricted actions or sensitive data. Bruce Schneier described prompt injection as an inherent, open-ended attack surface—ranging from obvious “ignore previous instructions” phrasing to more indirect techniques such as embedding malicious instructions in ASCII art or in text rendered inside images—arguing that point fixes for individual tricks do not provide universal protection with current LLM approaches.
New academic work highlighted that the same class of failures can extend into the physical world for embodied AI (e.g., robots, autonomous vehicles) through “environmental indirect prompt injection,” where misleading text placed on signs, posters, or objects is ingested by vision-language perception systems and treated as actionable instructions, potentially influencing real-world behavior. A separate TechXplore piece focused more broadly on responsible/ethical AI practices in industry (an interview format) and did not materially add technical detail on prompt injection or a specific security incident, making it less relevant to the core story about prompt-injection-driven hijacking risks and emerging attack pathways.
Get ahead of threats like this
Mallory correlates global threat intelligence with your attack surface — know if you’re exposed before adversaries strike.
How this story unfolded
4 events from the most recent confirmed update back to the earliest known activity.
Schneier highlights persistent prompt-injection weakness in AI systems
Bruce Schneier published an analysis arguing that LLMs remain fundamentally vulnerable to prompt injection because they flatten context into token patterns and lack human-like situational judgment. He warned that the risk is greater for autonomous AI agents using tools, and recommended narrow scoping and human escalation for high-risk uses.
Researchers publish arXiv preprint and announce SaTML 2026 presentation
The UC Santa Cruz study was made available as an arXiv preprint and slated for presentation at SaTML 2026. The authors also proposed defenses such as authenticating perceived text instructions and checking them against mission and safety constraints.
CHAI tests show physical-world text can steer robots, drones, and cars unsafely
The team evaluated CHAI in simulators, real driving photos, and a small robotic car in a university building, showing that printed text could override navigation and trigger unsafe outcomes such as crashes or inappropriate landings. Reported success rates reached 95.5% for aerial object tracking, 81.8% for driverless cars, and 68.1% for drone landing against systems using GPT-4o and InternVL.
UC Santa Cruz researchers develop CHAI prompt-injection attack for embodied AI
Researchers created CHAI (Command Hijacking against embodied AI), a two-stage attack pipeline that optimizes malicious real-world text and its physical presentation to manipulate visual-language-model-driven robots and vehicles. The work targeted embodied AI systems such as self-driving cars and drones.
Sources
2 references tracked. Mallory keeps watching after this page renders.
See the full picture, correlated to your attack surface.
Map indicators from this story to your assets and identify affected systems in minutes.
Every observed campaign, victim, and pivot linked to actors named in this story.
Malware, exploits, and IOCs connected to the activity described here.
YARA, Sigma, and Snort rules deployed to your SIEM as soon as they’re published.
Get matching new stories delivered to your team as they break — not the next morning.
Ask questions about this story and take action on the answers.
Related stories
Other stories Mallory is tracking that overlap on entities, sources, or topic labels.
Prompt injection and multimodal 'promptware' attacks against LLM-based systems
Security researchers and commentators warned that attacks on **LLM-based systems** are evolving beyond simple “prompt injection” into a broader execution mechanism dubbed **promptware**, with a proposed seven-step **promptware kill chain** to describe how malicious instructions enter and propagate through AI-enabled applications. The core risk highlighted is architectural: LLMs treat system instructions, user input, and retrieved content as a single token stream, enabling **indirect prompt injection** where hostile instructions are embedded in external data sources (web pages, emails, shared documents) that an LLM ingests at inference time; the attack surface expands further as models become **multimodal**, allowing instructions to be hidden in images or audio. Related academic work demonstrated a concrete multimodal variant against **embodied AI** using large vision-language models: **CHAI (Command Hijacking Against Embodied AI)**, which embeds deceptive natural-language instructions into visual inputs (e.g., road signs) to influence agent behavior in scenarios including drone emergency landing, autonomous driving, and object tracking, reportedly outperforming prior attacks in evaluations. Separately, reporting on a viral “AI caricature” social-media trend framed the risk as downstream **social engineering** and potential **LLM account takeover** leading to exposure of prompt histories and employer-sensitive data; while largely hypothetical, it underscores how widespread consumer LLM use and public oversharing can increase the likelihood and impact of prompt-driven compromise paths.
Mar 21, 2026
Prompt Injection Attacks and Security Challenges in AI Systems
Prompt injection has emerged as a critical security concern in the deployment of large language models (LLMs) and AI agents, with attackers exploiting the way these systems interpret and execute instructions. Security researchers have drawn parallels between prompt injection and earlier vulnerabilities like SQL injection, highlighting its potential to undermine the intended behavior of AI models. Prompt injection involves manipulating the input prompts to override or bypass the system-level instructions set by developers, leading to unauthorized actions or data leakage. The attack surface is broad, as LLMs are increasingly integrated into applications and workflows, making them attractive targets for adversaries. Multiple organizations, including OpenAI, Microsoft, and Anthropic, have initiated efforts to address prompt injection, but the problem remains unsolved due to the complexity and adaptability of AI models. Real-world demonstrations have shown that prompt injection can be used to break out of agentic applications, bypass browser security rules, and even persistently compromise AI systems through mechanisms like memory manipulation. Security conferences such as BlackHat USA 2024 have featured research on exploiting AI-powered tools like Microsoft 365 Copilot, where attackers can escalate privileges or exfiltrate data by crafting malicious prompts or leveraging markdown image vectors. Researchers have also identified that AI agents can be tricked into ignoring browser security policies, such as CORS, leading to potential cross-origin data leaks. Defensive measures, such as intentionally limiting AI capabilities or implementing stricter input filtering, have been adopted by some vendors, but these often come at the cost of reduced functionality. The security community is actively developing standards, such as the OWASP Agent Observability Standard, to improve monitoring and detection of prompt injection attempts. Despite these efforts, adversaries continue to find novel ways to exploit prompt injection, including dynamic manipulation of tool descriptions and bypassing image filtering mechanisms. The rapid evolution of AI technologies and the proliferation of agentic applications have made it challenging to keep pace with emerging threats. Security researchers emphasize the need for ongoing vigilance, robust testing, and collaboration across the industry to mitigate the risks associated with prompt injection. The use of AI in sensitive environments, such as enterprise productivity suites and web browsers, amplifies the potential impact of successful attacks. As AI adoption accelerates, organizations must prioritize understanding and defending against prompt injection to safeguard their systems and data. The ongoing research and public disclosures serve as a call to action for both developers and defenders to address this evolving threat landscape.
Mar 21, 2026
Prompt Injection Risks in Agentic AI and AI-Powered Browsers
Security researchers reported that **prompt injection** is enabling practical attacks against *agentic AI* systems that have access to tools and user data, and argued the industry is underestimating the threat. A proposed framing, **“promptware,”** describes malicious prompts as a malware-like execution mechanism that can drive an LLM to take actions via its connected tools—potentially leading to **data exfiltration**, cross-system propagation, IoT manipulation, or even **arbitrary code execution**, depending on the permissions and integrations available. Trail of Bits disclosed results from an adversarial security assessment of Perplexity’s *Comet* browser, showing how prompt injection techniques could be used to **extract private information from authenticated sessions (e.g., Gmail)** by abusing the browser’s AI assistant and its tool access (such as reading page content, using browsing history, and interacting with the browser). Their threat-model-driven testing emphasized that agentic assistants can treat external web content as instructions unless it is explicitly handled as **untrusted input**, and they published recommendations intended to reduce prompt-injection-driven data paths between the user’s local trust zone (profiles/cookies/history) and vendor-hosted agent/chat services.
Mar 21, 2026