// Intelligence Brief · Week 46

10 items that need action

2026-11-09 → 2026-11-15 UTC · 10 items selected
Correction · 2026-11-21 · writer_overreach
Correction, 21 November 2026: The first item in this edition (dnspython CVE-2026-128089, item ce2fd1aa) said services using dnspython for DNS resolution can be denied service, and told readers to upgrade wherever the package is a runtime dependency. The advisory is narrower: the from_text() parser can consume excessive memory and CPU when it reads a WKS record in attacker-controlled zone-file text. Applications are affected only when they parse zone-file text whose record type and rdata are attacker-controlled. The high severity rating and the patched version, 2.9.2, were stated correctly. This edition's other items are unchanged by this note.
The edition below is unchanged, except that corrected items show their corrected text with the original struck (ADR 0085); the record and the email as sent still carry the text as sent. ADR 0040: an edition is written once, so a correction sits beside the record rather than inside it — every correction, with its cause.
ACTION NEEDED PATCH_NOW
GitHub advisories — pip · 2026-11-15 06:36 UTC · Model-reported confidence 85% (not measured accuracy)
Event date: unknown; not publication time
Affected audience: Security and platform owners of the affected stack
WHAT CHANGEDGitHub published GHSA-vvf4-9x63-8xvv on 13 November 2026, rating CVE-2026-128089 in dnspython as high severity. The vulnerability is in zone-file parsing: the `from_text()` parser does not limit a WKS record service number to 0–65535, so a maliciously crafted zone file can cause excessive memory and CPU consumption. Versions earlier than 2.9.2 are affected. First patched version is 2.9.2. WHY IT MATTERSApplications are affected only when they parse zone-file text whose record type and rdata are attacker-controlled. Using dnspython only for DNS resolution, without parsing attacker-controlled zone-file text, does not match the advisory's stated impact. A high severity rating does not mean every deployment that includes the package is exposed. WHAT TO DOIf your code parses zone-file format with dnspython (including `dns.rdata.from_text`) where an attacker can control the record type and rdata, upgrade to version 2.9.2 or later. If you only use dnspython as a DNS resolver and do not parse untrusted zone-file text, this advisory does not describe a demonstrated exposure, though upgrading remains ordinary hygiene.
Original text (as first published, superseded)
WHAT CHANGEDGitHub published GHSA-vvf4-9x63-8xvv on 13 November 2026, rating CVE-2026-128089 in dnspython as high severity. The vulnerability causes excessive memory and CPU consumption. First patched version is 2.9.2.WHY IT MATTERSServices using dnspython for DNS resolution can experience availability degradation or denial of service through resource exhaustion. The high severity rating indicates exploitability affects production workloads.WHAT TO DOUpgrade dnspython to version 2.9.2 or later in services where it is a runtime dependency.
Full item and evidence ↗  ·  Primary source ↗
ACTION NEEDED PATCH_NOW
GitHub advisories — pip · 2026-11-15 06:34 UTC · Model-reported confidence 85% (not measured accuracy)
Event date: unknown; not publication time
Affected audience: Security and platform owners of the affected stack
WHAT CHANGEDGitHub published GHSA-6655-6qjp-jx7x on 13 November 2026, a high-severity advisory for CVE-2026-126191 in soupsieve. The vulnerability is an exponential regular expression denial of service (ReDoS) in selector compilation when processing \9-run sequences in CSS_ESCAPES/IDENTIFIER. soupsieve 3.0.2 patches the issue. WHY IT MATTERSsoupsieve is a CSS selector library used throughout Python dependency trees, including by BeautifulSoup. Malicious or malformed CSS selectors can cause exponential processing time, degrading availability in applications that parse untrusted HTML or CSS. WHAT TO DOCheck whether your dependencies include soupsieve versions before 3.0.2. If so, update to 3.0.2 or later in this cycle.
Full item and evidence ↗  ·  Primary source ↗
ACTION NEEDED MITIGATE
CISA Known Exploited Vulnerabilities · 2026-11-15 06:33 UTC · Model-reported confidence 85% (not measured accuracy)
Event date: unknown; not publication time
Affected audience: Security and platform owners of the affected stack
WHAT CHANGEDCISA added CVE-2026-123039 to the Known Exploited Vulnerabilities catalog on 2026-11-13. The vulnerability affects multiple Check Point products and is classified as an incorrect permission assignment for critical resource. Due date for federal agencies is 2026-11-16. WHY IT MATTERSActive exploitation is confirmed. Permission assignment flaws typically allow attackers to escalate privileges or move laterally once inside the network. Federal agencies and contractors must comply with BOD 26-04. WHAT TO DOCheck your Check Point deployments against vendor guidance for CVE-2026-123039. Apply vendor mitigations by 2026-11-16 if you are subject to BOD 26-04, or discontinue use if mitigations are unavailable.
Full item and evidence ↗  ·  Primary source ↗
ACTION NEEDED VERIFY_EXPOSURE
CISA Known Exploited Vulnerabilities · 2026-11-15 06:33 UTC · Model-reported confidence 85% (not measured accuracy)
Event date: unknown; not publication time
Affected audience: Security and platform owners of the affected stack
WHAT CHANGEDCISA added CVE-2026-114713, an improper neutralization of delimiters vulnerability in UnrealIRCd, to the Known Exploited Vulnerabilities catalog on 2026-11-13. The vulnerability is being actively exploited in the wild. Due date for federal agencies is 2026-11-16. WHY IT MATTERSActive exploitation of this IRC daemon vulnerability means attackers have working exploits. Federal agencies must act by 2026-11-16 under BOD 26-04. Private sector organizations running UnrealIRCd face the same threat with no compliance deadline protecting them from delayed action. WHAT TO DOCheck whether you run UnrealIRCd in any environment. If yes, apply mitigations per vendor instructions immediately. Remediation: see bulletin https://www.cisa.gov/known-exploited-vulnerabilities-catalog
Full item and evidence ↗  ·  Primary source ↗
ACTION NEEDED BUDGET
OpenRouter model catalogue · 2026-11-15 06:32 UTC · Model-reported confidence 95% (not measured accuracy)
Event date: unknown; not publication time
Affected audience: Engineering and product teams using hosted AI APIs
WHAT CHANGEDOpenRouter's captured prompt base rate for deepseek/deepseek-v4.1-flash-1017 changed from $0.006 to $0.038 per 1M tokens. Completion pricing remains $0.50 per 1M tokens. WHY IT MATTERSUsing the captured base rates for deepseek/deepseek-v4.1-flash-1017, 100M prompt tokens per day cost $0.60 before and $3.80 after, excluding all other charges. To estimate the added prompt cost for deepseek/deepseek-v4.1-flash-1017 in dollars, multiply prompt token volume by 0.032 and divide by 1,000,000, using the captured base-rate difference in dollars per 1M tokens. WHAT TO DOReforecast costs for any workloads using deepseek/deepseek-v4.1-flash-1017 through OpenRouter. Review current month-to-date token volumes and update budget projections using the new $0.038/M prompt rate.
Full item and evidence ↗  ·  Primary source ↗
ACTION NEEDED VERIFY_ONLY
Claude Code releases · 2026-11-15 06:34 UTC · Model-reported confidence 85% (not measured accuracy)
Event date: 2026-11-13
Affected audience: Developers and operators building agent applications
WHAT CHANGEDClaude Code v2.1.325, released 13 November 2026, changed sandboxed Bash commands to follow the normal permission flow by default instead of running without prompts. The previous auto-allow behavior is now opt-in via sandbox.autoAllowBashIfSandboxed: true. The release also fixed a security issue in zstd-compressed response decoding, changed read-only git commands to require permission in untrusted repositories, and changed scheduled tasks and /loop to ask before running skills that declare allowed-tools or hooks. WHY IT MATTERSExisting workflows that relied on sandboxed commands running automatically will now pause for permission prompts. Scheduled tasks and automation that depended on the previous behavior will stop executing until the setting is added or permissions are granted interactively. WHAT TO DOTest scheduled tasks, /loop workflows, and any automation that runs sandboxed Bash commands. If they now block on permission prompts, add sandbox.autoAllowBashIfSandboxed: true to your managed settings or add Skill(name) to permissions.allow for affected skills. For git commands in untrusted repositories, use --add-dir instead of additionalDirectories to avoid prompts.
Full item and evidence ↗  ·  Primary source ↗
ACTION NEEDED MIGRATE
OpenAI deprecations · 2026-11-13 06:50 UTC · Model-reported confidence 95% (not measured accuracy)
Event date: unknown; not publication time
Affected audience: Engineering and product teams using hosted AI APIs
WHAT CHANGEDOpenAI deprecated gpt-image-2 and gpt-image-2-2026-04-21 on 2026-11-10, with removal from the API on February 10, 2027. The deprecation policy now explicitly states image generation models receive at least 3 months notice, alongside specialized variants of generally available models. WHY IT MATTERSAny application using gpt-image-2 or its April 2026 snapshot will break after February 10, 2027. The 3-month notice window for image models is now formalized, establishing the minimum migration time for future image model retirements. WHAT TO DOGrep codebases for 'gpt-image-2' and 'gpt-image-2-2026-04-21'. Migrate to gpt-image-2.6-sunburst or gpt-image-2.6-flare before February 10, 2027.
Full item and evidence ↗  ·  Primary source ↗
ACTION NEEDED MIGRATE
Cloudflare developer changelog · 2026-11-15 06:34 UTC · Model-reported confidence 95% (not measured accuracy)
Event date: 2026-11-13
Affected audience: Developers and operators building agent applications
WHAT CHANGEDCloudflare Terraform provider v5.29.0 changes the cloudflare_ai_search_instance import ID format from <account_id>/<id> to <account_id>/<namespace>/<id>. The cloudflare_ai_search_instance and cloudflare_ai_search_instances data sources now require a namespace argument. The filter.namespace argument is removed from cloudflare_ai_search_instances. The cloudflare_ai_search_instance resource adds an optional namespace argument that defaults to "default". WHY IT MATTERSExisting Terraform configurations that import AI Search instances or use AI Search data sources will fail after upgrading to v5.29.0 unless the import commands and data source blocks are updated to include the namespace parameter. WHAT TO DOBefore upgrading to Terraform provider v5.29.0, update import commands to the format terraform import cloudflare_ai_search_instance.name <account_id>/default/<id> for default namespace instances. Add namespace = "default" to all cloudflare_ai_search_instance and cloudflare_ai_search_instances data source blocks. Remove any filter.namespace arguments from cloudflare_ai_search_instances data sources.
Full item and evidence ↗  ·  Primary source ↗
ACTION NEEDED VERIFY_EXPOSURE
AWS security bulletins · 2026-11-14 18:03 UTC · Deterministic source evidence; confidence not scored
Event date: 2026-11-12
Affected audience: Security and platform owners of the affected stack
WHAT CHANGEDCVE-2026-126793 - Privilege escalation in Aurora PostgreSQL dialect detection in AWS Advanced Go Wrapper. Source event: 2026-11-12T21:11:28+00:00. RSS discovery: 2026-11-13T06:45:29.599887+00:00. WHY IT MATTERSThe captured feed identifies CVE-2026-126793. Check exposure against the affected product and versions; detailed remediation evidence is unavailable. WHAT TO DOCheck your inventory for the product and affected versions named in this advisory. Remediation details were not captured; consult the vendor bulletin before changing configuration or applying a fix. Remediation: see bulletin https://aws.amazon.com/security/security-bulletins/2026-182-aws/ (retrieval attempted 2026-11-14T18:03:56.002Z; RSS discovery 2026-11-13T06:45:29.599887+00:00).
Full item and evidence ↗  ·  Primary source ↗
ACTION NEEDED VERIFY_EXPOSURE
AWS security bulletins · 2026-11-14 18:03 UTC · Deterministic source evidence; confidence not scored
Event date: 2026-11-12
Affected audience: Security and platform owners of the affected stack
WHAT CHANGEDCVE-2026-127119 - Unverified ownership of the default EMR cluster log location in AWS SDK for pandas. Source event: 2026-11-12T19:07:08+00:00. RSS discovery: 2026-11-13T06:45:29.493892+00:00. WHY IT MATTERSThe captured feed identifies CVE-2026-127119. Check exposure against the affected product and versions; detailed remediation evidence is unavailable. WHAT TO DOCheck your inventory for the product and affected versions named in this advisory. Remediation details were not captured; consult the vendor bulletin before changing configuration or applying a fix. Remediation: see bulletin https://aws.amazon.com/security/security-bulletins/2026-181-aws/ (retrieval attempted 2026-11-14T18:03:51.431Z; RSS discovery 2026-11-13T06:45:29.493892+00:00).
Full item and evidence ↗  ·  Primary source ↗

What the week cost

20runs
$4.2039model cost
40sources watched
0corrections
Printed because they are checkable. No item published this week has been found wrong. That is not a claim of accuracy — how precision is counted.
RADAR watches 40 sources every twelve hours (scheduled runs often start hours late) and writes up only what changed materially. Get it weekly or daily by email — or take the feed, which is free and has terms that say so.
First published 2026-11-16 22:24 UTC. The items, their order and the numbers above are frozen as of that moment and are never regenerated — every published edition.