ryan ξ¨ retweeted
π¨ BREAKING: Google DeepMind just mapped the attack surface that nobody in AI is talking about.
Websites can already detect when an AI agent visits and serve it completely different content than humans see.
> Hidden instructions in HTML.
> Malicious commands in image pixels.
> Jailbreaks embedded in PDFs.
Your AI agent is being manipulated right now and you can't see it happening.
The study is the largest empirical measurement of AI manipulation ever conducted. 502 real participants across 8 countries.
23 different attack types. Frontier models including GPT-4o, Claude, and Gemini.
The core finding is not that manipulation is theoretically possible it is that manipulation is already happening at scale and the defenses that exist today fail in ways that are both predictable and invisible to the humans who deployed the agents.
Google DeepMind built a taxonomy of every known attack vector, tested them systematically, and measured exactly how often they work.
The results should alarm everyone building agentic systems.
The attack surface is larger than anyone has publicly acknowledged. Prompt injection where malicious instructions hidden in web content hijack an agent's behavior works through at least a dozen distinct channels.
Text hidden in HTML comments that humans never see but agents read and follow. Instructions embedded in image metadata.
Commands encoded in the pixels of images using steganography, invisible to human eyes but readable by vision-capable models.
Malicious content in PDFs that appears as normal document text to the agent but contains override instructions.
QR codes that redirect agents to attacker-controlled content.
Indirect injection through search results, calendar invites, email bodies, and API responses any data source the agent consumes becomes a potential attack vector.
The detection asymmetry is the finding that closes the escape hatch. Websites can already fingerprint AI agents with high reliability using timing analysis, behavioral patterns, and user-agent strings.
This means the attack can be conditional: serve normal content to humans, serve manipulated content to agents.
A user who asks their AI agent to book a flight, research a product, or summarize a document has no way to verify that the content the agent received matches what a human would see.
The agent cannot tell the user it was served different content.
It does not know. It processes whatever it receives and acts accordingly.
The attack categories and what they enable:
β Direct prompt injection: malicious instructions in any text the agent reads overrides goals, exfiltrates data, triggers unintended actions
β Indirect injection via web content: hidden HTML, CSS visibility tricks, white text on white backgrounds invisible to humans, consumed by agents
β Multimodal injection: commands in image pixels via steganography, instructions in image alt-text and metadata
β Document injection: PDF content, spreadsheet cells, presentation speaker notes every file format is a potential vector
β Environment manipulation: fake UI elements rendered only for agent vision models, misleading CAPTCHA-style challenges
β Jailbreak embedding: safety bypass instructions hidden inside otherwise legitimate-looking content
β Memory poisoning: injecting false information into agent memory systems that persists across sessions
β Goal hijacking: gradual instruction drift across multiple interactions that redirects agent objectives without triggering safety filters
β Exfiltration attacks: agents tricked into sending user data to attacker-controlled endpoints via legitimate-looking API calls
β Cross-agent injection: compromised agents injecting malicious instructions into other agents in multi-agent pipelines
The defense landscape is the most sobering part of the report.
Input sanitization cleaning content before the agent processes it fails because the attack surface is too large and too varied.
You cannot sanitize image pixels. You cannot reliably detect steganographic content at inference time.
Prompt-level defenses that tell agents to ignore suspicious instructions fail because the injected content is designed to look legitimate.
Sandboxing reduces the blast radius but does not prevent the injection itself. Human oversight the most commonly cited mitigation fails at the scale and speed at which agentic systems operate.
A user who deploys an agent to browse 50 websites and summarize findings cannot review every page the agent visited for hidden instructions.
The multi-agent cascade risk is where this becomes a systemic problem.
In a pipeline where Agent A retrieves web content, Agent B processes it, and Agent C executes actions, a successful injection into Agent A's data feed propagates through the entire system.
Agent B has no reason to distrust content that came from Agent A. Agent C has no reason to distrust instructions that came from Agent B.
The injected command travels through the pipeline with the same trust level as legitimate instructions. Google DeepMind documents this explicitly: the attack does not need to compromise the model.
It needs to compromise the data the model consumes. Every agentic system that reads external content is one carefully crafted webpage away from executing attacker instructions.
The agents are already deployed. The attack infrastructure is already being built. The defenses are not ready.
ryan ξ¨ retweeted
The Iranian attack on Prince Sultan Air Base in Saudi Arabia was apparently bigger than first thought, with reports of damage to multiple refueling aircraft, an E-3 Sentry AWACS, and injuries to at least 10 U.S. troops.
Updated report: theaviationist.com/2026/03/2β¦
ryan ξ¨ retweeted
BREAKING: According to this satellite image taken by Landsat 8/9, it appears that, in addition to the KC-135R Stratotankers of the U.S. Air Force, at least one or two E-3G Sentry AWACS has also been destroyed in the ballistic missile strike carried out by the IRGC Aerospace Force at Prince Sultan Air Base, Saudi Arabia, on March 27, 2026.
In fact, it appears that, the primary target of the IRGCASF was the E-3Gs of the USAF parked there, not the KC-135Rs. However, they also managed to destroy several KC-135Rs.
What makes this even more personal for me is that I had previously met and photographed those E-3G Sentry AWACS aircraft weeks before start of this war for my book on Op. Epic Fury.
#OperationEpicFury #OperationLionsRoar
ryan ξ¨ retweeted
INTERESTING: A surveillance drone, reportedly flown by infiltrator elements and disguised as a natural bird β such as an eagle β has been spotted in Iran.
ryan ξ¨ retweeted
Introducing TurboQuant: Our new compression algorithm that reduces LLM key-value cache memory by at least 6x and delivers up to 8x speedup, all with zero accuracy loss, redefining AI efficiency. Read the blog to learn how it achieves these results: goo.gle/4bsq2qI
ryan ξ¨ retweeted
LiteLLM HAS BEEN COMPROMISED, DO NOT UPDATE. We just discovered that LiteLLM pypi release 1.82.8. It has been compromised, it contains litellm_init.pth with base64 encoded instructions to send all the credentials it can find to remote server + self-replicate. link below
ryan ξ¨ retweeted
A significant movement is underway from US Army, Navy and Air Force bases in CONUS to the Middle East comprised of at least 35 C-17 flights since March 12th, with 11 more flights on the way.
Origins:
12-Hunter Army Air Field/Fort Stewart, GA
8-Unknown
7-JB Lewis-McChord, WA
6-Pope Army Air Field/Fort Bragg, NC
4-Campbell Army Airfield/Fort Campbell, KY
4-Gray Army Airfield/JB Lewis-McChord, WA
4-Naval Air Station Oceana, VA
1-MacDill AFB, FL
1-JB McGuire-Dix-Lakehurst, NJ
Destinations:
17-Ovda Air Base, Israel
13-King Faisal Air Base, Jordan
4-King Hussein Int'l Airport, Jordan
ryan ξ¨ retweeted
Deployment of the 82nd Airborne Division from Ft. Bragg, as well as other units with the U.S. Army, appear to now be underway from bases within the Continental United States towards the Middle East, as staging continues for potential ground operations in Iran.
A significant movement is underway from US Army, Navy and Air Force bases in CONUS to the Middle East comprised of at least 35 C-17 flights since March 12th, with 11 more flights on the way.
Origins:
12-Hunter Army Air Field/Fort Stewart, GA
8-Unknown
7-JB Lewis-McChord, WA
6-Pope Army Air Field/Fort Bragg, NC
4-Campbell Army Airfield/Fort Campbell, KY
4-Gray Army Airfield/JB Lewis-McChord, WA
4-Naval Air Station Oceana, VA
1-MacDill AFB, FL
1-JB McGuire-Dix-Lakehurst, NJ
Destinations:
17-Ovda Air Base, Israel
13-King Faisal Air Base, Jordan
4-King Hussein Int'l Airport, Jordan
ryan ξ¨ retweeted
btw their supabase storage bucket is publicly accessible via any signed url token π
exposes:
> employee background checks
> equity vesting schedules and grant amounts
> performance reviews
> session tokens for stripe, notion, etc
> screenshots below π§΅
i also got access to their notion π
A detailed and brutal look at the tactics of buzzy AI compliance startup Delve
"Delve built a machine designed to make clients complicit without their knowledge, to manufacture plausible deniability while producing exactly the opposite."
substack.com/home/post/p-191β¦
ryan ξ¨ retweeted
IRAQ ANNOUNCES A FORCE MAJEURE ON ALL OILFIELDS OPERATED BY FOREIGN OIL COMPANIES, ACCORDING TO OIL MINISTRY SOURCES.
No shit
A widespread transition to renewable energy could mitigate a major cause of international conflict in a future that moves away from fossil fuels, energy and climate change experts told ABC News.
Read more: abcnews.link/Jb4QZLu
1
ryan ξ¨ retweeted
That's a strong statement from Lloydβs of London's head of underwriting : β[we] still provide cover to basically anyone who asksβ and the flow through Hormuz is βnot an insurance issueβ
ryan ξ¨ retweeted
Merchant vessel Seaway Hawk transporting decommissioned U.S. Navy Avenger-class Mine Countermeasures Ships USS Devastator, USS Dextrous, USS Gladiator and USS Sentry in the Delaware Bay to unload in Philadelphia tomorrow - March 9, 2026 SRC: FB- Delaware River Shipping / Andy Wyrick π·


































