The authors constructed images that looked largely ordinary at 512-pixel scale but made injected instructions more legible after downsampling to a 64-pixel input. The exploit depends on how the agent preprocesses pixels, not merely on whether a human would notice suspicious text in the source image. Across selected model tests, text recognition was much…
Inspect the ClaimsCase · DiggingBeagle record
Scaling an image could reveal attacker text to a vision-language model that was inconspicuous at the original resolution
The authors constructed images that looked largely ordinary at 512-pixel scale but made injected instructions more legible after downsampling to a 64-pixel input. The exploit depends on how the agent preprocesses pixels, not merely on whether a human would notice suspicious text in the source image.
Public disclosure: Nov 12, 2026
An agent retrieves or opens an attacker-influenced image and preprocesses it at a resolution different from the human-inspected artifact.
At a glance
Untrusted image pixels cross into the model instruction channel when text only becomes apparent after resizing or enhancement.
Read the implicationsThe evidence is bounded by the cited source and test setting. Broader incidence, other environments, and actual production consequences are not established merely by these observations.
Limits and uncertaintyFull account
The authors constructed images that looked largely ordinary at 512-pixel scale but made injected instructions more legible after downsampling to a 64-pixel input. The exploit depends on how the agent preprocesses pixels, not merely on whether a human would notice suspicious text in the source image.
Across selected model tests, text recognition was much higher than the rate of successfully induced tool actions. In the web-agent evaluation, attacker-relevant behavior rose to 4.00 percent for Gemini-3-Flash and 3.67 percent for GPT-4o, despite much higher perception of injected instructions.
The reported images are a controlled cross-resolution prompt-injection demonstration, not proof that most multimodal agents obey such instructions or that a deployed customer was compromised.
Reported figures and denominator. Selected web-agent tool-behavior success 4.00% and 3.67%, not the 54% to 63% visual recognition figures.
The evidence is bounded by the cited source and test setting. Broader incidence, other environments, and actual production consequences are not established merely by these observations.
Mechanism and trust boundary
- 01
Starting condition
An agent retrieves or opens an attacker-influenced image and preprocesses it at a resolution different from the human-inspected artifact.
Boundary: precondition and attacker capability
- 02
Trust boundary crossed or evaluated
Across selected model tests, text recognition was much higher than the rate of successfully induced tool actions. In the web-agent evaluation, attacker-relevant behavior rose to 4.00 percent for Gemini-3-Flash and 3.67 percent for GPT-4o, despite much higher perception of injected instructions.
Boundary: mechanism
- 03
Observed consequence
Untrusted image pixels cross into the model instruction channel when text only becomes apparent after resizing or enhancement.
Boundary: observed effect and limitations
Claims & evidence
CLM-NOV15-SISS-MECHThe authors constructed images that looked largely ordinary at 512-pixel scale but made injected instructions more legible after downsampling to a 64-pixel input. The exploit depends on how the agent preprocesses pixels, not merely on whether a human would notice suspicious text in the source image. Across selected model tests, text recognition was much higher than the rate of successfully induced tool actions. In the web-agent evaluation, attacker-relevant behavior rose to 4.00 percent for Gemini-3-Flash and 3.67 percent for GPT-4o, despite much higher perception of injected instructions.supported
Basis: reported finding
- supportsScale-Induced Semantic Shift in Visual Prompt Injectionprimary disclosure
SISS paper, image scaling setup and Table II web-agent behavior vs perception
CLM-NOV15-SISS-SCOPEThe reported images are a controlled cross-resolution prompt-injection demonstration, not proof that most multimodal agents obey such instructions or that a deployed customer was compromised. The evidence is bounded by the cited source and test setting. Broader incidence, other environments, and actual production consequences are not established merely by these observations.supported
Basis: reported finding
- supportsScale-Induced Semantic Shift in Visual Prompt Injectionprimary disclosure
Primary disclosure: results, scope and limitations
Implications
Untrusted image pixels cross into the model instruction channel when text only becomes apparent after resizing or enhancement.
Controls and mitigations
- Inspect inputs at the actual model resolution and at multiple nearby scales before granting authority to extracted text.
- Require external authorization for mutating tool calls regardless of whether an image instruction is recognized; evaluate real side effects separately from text-recognition rate.
Unknowns and contradictions
- The evidence is bounded by the cited source and test setting. Broader incidence, other environments, and actual production consequences are not established merely by these observations.
Sources and citation
Material revision history
- Nov 15, 2026 · Canonical change recorded · new in release · revision 250