VAITP Dataset

← Back to the dataset

CVE-2026-107291

OpenTelemetry can leak sensitive agent content despite include_content=False setting.

  • CVSS 2.3
  • CWE-212
  • Information Leakage
  • Remote

Pydantic AI is a Python agent framework for building applications and workflows with Generative AI. From 0.3.4 until 1.107.6 and 2.44.0, OpenTelemetry instrumentation configured with InstrumentationSettings(include_content=False) can still export sensitive agent content through exception.message and exception.stacktrace events, error status descriptions, and model_request_parameters containing instructions or the prompted_output_template. The exposed data is available to readers of the configured telemetry backend and can include tool feedback, provider error bodies, runtime instructions, and structured-output templates even though message attributes are redacted. This issue does not grant new access to agent data, and deployments that do not use include_content=False are not affected by the setting bypass. This issue is fixed in versions 1.107.6 and 2.44.0.

CVSS base score
2.3
Published
2026-10-08
OWASP
A09 Security Logging and Monitoring Failures
Orthogonal defect classification
Interface
Code defect classification
Incorrect Interface
Category
Information Leakage
Subcategory
Information Disclosure
Accessibility scope
Remote
Impact
Information Disclosure
Affected component
Pydantic AI
Fixed by upgrading
Yes

Solution

Upgrade to Pydantic AI v1.107.6 or later (or v2.44.0 or later).

Vulnerable code sample

from opentelemetry import trace
from opentelemetry.sdk.trace import TracerProvider
from opentelemetry.sdk.trace.export import SimpleSpanProcessor, ConsoleSpanExporter
from pydantic_ai.instrumentation import InstrumentationSettings, PydanticAIOtelInstrumentation

# Setup tracing
trace.set_tracer_provider(TracerProvider())
trace.get_tracer_provider().add_span_processor(
    SimpleSpanProcessor(ConsoleSpanExporter())
)

# VULNERABLE: exception details are added despite include_content=False
instrumentation = PydanticAIOtelInstrumentation(
    settings=InstrumentationSettings(include_content=False)
)
instrumentation.instrument()

def run_agent():
    tracer = trace.get_tracer(__name__)
    with tracer.start_as_current_span("agent.run") as span:
        try:
            # Simulate agent error that includes sensitive instruction
            raise ValueError("Secret instruction: DELETE_ALL_DATA")
        except Exception as e:
            # Leak sensitive message and stacktrace
            span.set_attribute("exception.message", str(e))
            span.set_attribute("exception.stacktrace", "".join(trace.format_exception(type(e), e, e.__traceback__)))
            raise

run_agent()

Patched code sample

from opentelemetry import trace
from opentelemetry.sdk.trace import TracerProvider
from opentelemetry.sdk.trace.export import SimpleSpanProcessor, ConsoleSpanExporter
from pydantic_ai.instrumentation import InstrumentationSettings, PydanticAIOtelInstrumentation

# Setup tracing
trace.set_tracer_provider(TracerProvider())
trace.get_tracer_provider().add_span_processor(
    SimpleSpanProcessor(ConsoleSpanExporter())
)

# FIX: avoid adding exception details when include_content=False
instrumentation = PydanticAIOtelInstrumentation(
    settings=InstrumentationSettings(include_content=False)
)
instrumentation.instrument()

def run_agent():
    tracer = trace.get_tracer(__name__)
    with tracer.start_as_current_span("agent.run") as span:
        try:
            # Simulate agent error that includes sensitive instruction
            raise ValueError("Secret instruction: DELETE_ALL_DATA")
        except Exception as e:
            # Do not expose exception.message or stacktrace
            span.record_exception(e)  # OpenTelemetry will respect include_content setting
            raise

run_agent()

Published fix

Public sources link this CVE to the changes below. The code samples above were generated to illustrate the defect; these links lead to the code the project actually changed.

  • Follow the same exception recording rule on the model request and realtime spans (#8408) (pydantic/pydantic-ai)
    pydantic_ai_slim/pydantic_ai/capabilities/instrumentation.py, pydantic_ai_slim/pydantic_ai/capabilities/instrumentation.py · Instrumentation.wrap_run, pydantic_ai_slim/pydantic_ai/realtime/_instrumentation.py, pydantic_ai_slim/pydantic_ai/realtime/_instrumentation.py · SessionInstrumentation.record_error, pydantic_ai_slim/pydantic_ai/realtime/_session.py · RealtimeSession.close, pydantic_ai_slim/pydantic_ai/_instrumentation.py, pydantic_ai_slim/pydantic_ai/_instrumentation.py · open_model_request_span
  • Share exception recording between agent run and tool spans in `Instrumentation` (#8403) (pydantic/pydantic-ai)
    pydantic_ai_slim/pydantic_ai/capabilities/instrumentation.py, pydantic_ai_slim/pydantic_ai/capabilities/instrumentation.py · Instrumentation.on_tool_validate_error, pydantic_ai_slim/pydantic_ai/capabilities/instrumentation.py · Instrumentation.wrap_run, pydantic_ai_slim/pydantic_ai/capabilities/instrumentation.py · Instrumentation._run_tool_span
  • Close two `include_content` redaction bypasses and make the content sweep prove itself (v1 backport) (#8429) (pydantic/pydantic-ai)
    pydantic_ai_slim/pydantic_ai/capabilities/instrumentation.py, pydantic_ai_slim/pydantic_ai/capabilities/instrumentation.py · Instrumentation.wrap_run, pydantic_ai_slim/pydantic_ai/models/fallback.py, pydantic_ai_slim/pydantic_ai/models/fallback.py · FallbackModel._set_span_attributes, pydantic_ai_slim/pydantic_ai/_instrumentation.py, pydantic_ai_slim/pydantic_ai/_instrumentation.py · model_request_parameters_attributes, pydantic_ai_slim/pydantic_ai/_instrumentation.py · open_model_request_span, pydantic_ai_slim/pydantic_ai/_instrumentation.py · record_exception, pydantic_ai_slim/pydantic_ai/_instrumentation.py · set_error_status
  • Share exception recording between the agent run, tool and model request spans in `Instrumentation` (v1 backport) (#8404) (pydantic/pydantic-ai)
    pydantic_ai_slim/pydantic_ai/capabilities/instrumentation.py, pydantic_ai_slim/pydantic_ai/capabilities/instrumentation.py · Instrumentation.wrap_run, pydantic_ai_slim/pydantic_ai/capabilities/instrumentation.py · Instrumentation._run_tool_span, pydantic_ai_slim/pydantic_ai/_instrumentation.py, pydantic_ai_slim/pydantic_ai/_instrumentation.py · open_model_request_span
  • Close two `include_content` redaction bypasses and make the content sweep prove itself (#8428) (pydantic/pydantic-ai)
    pydantic_ai_slim/pydantic_ai/models/fallback.py, pydantic_ai_slim/pydantic_ai/models/fallback.py · FallbackModel._set_span_attributes, pydantic_ai_slim/pydantic_ai/_instrumentation.py, pydantic_ai_slim/pydantic_ai/_instrumentation.py · capture_current_context, pydantic_ai_slim/pydantic_ai/_instrumentation.py · model_request_parameters_attributes, pydantic_ai_slim/pydantic_ai/_instrumentation.py · open_model_request_span, pydantic_ai_slim/pydantic_ai/_instrumentation.py · _serialize_model_request_parameters

Cite this entry

@misc{vaitp:cve2026107291,
  title        = {{OpenTelemetry can leak sensitive agent content despite include_content=False setting.}},
  author       = {Bogaerts, Fr\'ed\'eric and Ivaki, Naghmeh and Fonseca, Jos\'e},
  year         = {2026},
  note         = {VAITP Python Vulnerability Dataset, entry CVE-2026-107291},
  howpublished = {\url{https://netpack.pt/vaitp/vulnerability/CVE-2026-107291/}}
}

Download this entry as JSON Download the whole dataset

Introducing the "VAITP dataset": a specialized repository of Python vulnerabilities and patches, meticulously compiled for the use of the security research community. As Python's prominence grows, understanding and addressing potential security vulnerabilities become crucial. Crafted by and for the cybersecurity community, this dataset offers a valuable resource for researchers, analysts, and developers to analyze and mitigate the security risks associated with Python. Through the comprehensive exploration of vulnerabilities and corresponding patches, the VAITP dataset fosters a safer and more resilient Python ecosystem, encouraging collaborative advancements in programming security.

The supreme art of war is to subdue the enemy without fighting.

Sun Tzu – “The Art of War”

:: Shaping the future through research and ingenuity ::

Before you use the VAITP AI models

Please read and accept the following disclaimer:

By choosing "I understand and accept", you acknowledge that you understand and agree to these terms.