What changed in AI security, Aug 3 to Aug 9, 2026
Aug 3 to Aug 9, 2026 (ISO week 2026-W32). Weeks run Monday to Sunday in UTC.
240 records published, +45 on the previous week: 79 vulnerabilities (+58), 2 incidents (+2), 3 research items (-2), 152 news items (-15), 4 policy items (+2).
Critical and high advisories
Vulnerability records rated critical or high, newest first. Showing 25 of 59.- Critical
CVE-2026-14526: The AI Copilot – Content Generator plugin for WordPress is vulnerable to authorization bypass in all versions up to…
CVE-2026-14526NVD/CVE Database - High
CVE-2026-12261: nltk.downloader cross-package resource and model poisoning via shared namespaces
CVE-2026-12261NVD/CVE Database - Critical
CVE-2026-67622: Flowise insecure direct object reference in OpenAI Assistants integration
CVE-2026-67622NVD/CVE Database - High
CVE-2026-19111 - Insecure direct object reference in Strands Agents Tools memory tools
AWS Security Bulletins - Critical
CVE-2026-67531: FrontMCP sandbox escape to remote code execution through codecall:execute tool
CVE-2026-67531NVD/CVE Database - High
CVE-2026-18954 - Incorrect authorization in the aggregation pipeline tool in Amazon AWS Labs DocumentDB MCP Server
AWS Security Bulletins - High
CVE-2026-18953 - Improper limitation of a pathname in AWS Transform MCP Server
AWS Security Bulletins - High
CVE-2026-69111: Milvus unauthenticated denial of service via /management/stop endpoint
CVE-2026-69111NVD/CVE Database - High
CVE-2026-9205: IBM Langflow OSS weak key derivation in ensure_fernet_key()
CVE-2026-9205NVD/CVE Database - High
CVE-2026-9201: IBM Langflow OSS code execution via truncated hash in component validation
CVE-2026-9201NVD/CVE Database - High
CVE-2026-9196: IBM Langflow OSS code execution during Agentic Assistant validation
CVE-2026-9196NVD/CVE Database - High
CVE-2026-9130: IBM Langflow OSS authorization bypass in MemoryComponent
CVE-2026-9130NVD/CVE Database - High
CVE-2026-8478: IBM Langflow OSS remote code injection through improper control of user input
CVE-2026-8478NVD/CVE Database - High
CVE-2026-8470: IBM Langflow OSS weak Fernet key generation from user secrets
CVE-2026-8470NVD/CVE Database - High
CVE-2026-8183: IBM Langflow OSS path traversal via crafted URL request
CVE-2026-8183NVD/CVE Database - High
CVE-2026-8182: IBM Langflow OSS unauthenticated remote code execution via HTTP requests
CVE-2026-8182NVD/CVE Database - Critical
CVE-2026-48168: PraisonAI command injection via pull request branch name
CVE-2026-48168NVD/CVE Database - High
CVE-2026-17633: IBM Langflow OSS remote code execution through code injection
CVE-2026-17633NVD/CVE Database - High
CVE-2026-17632: IBM Langflow OSS remote code execution through AST-based security scanning
CVE-2026-17632NVD/CVE Database - High
CVE-2026-17624: IBM Langflow OSS code execution through improper module import validation
CVE-2026-17624NVD/CVE Database - High
CVE-2026-9081: IBM Langflow OSS SSRF in Ollama provider key validation
CVE-2026-9081NVD/CVE Database - High
CVE-2026-17625: IBM Langflow OSS OS command injection by remote authenticated attacker
CVE-2026-17625NVD/CVE Database - High
CVE-2026-9077: IBM Langflow OSS localhost restriction bypass via remote authenticated access
CVE-2026-9077NVD/CVE Database - High
CVE-2026-8446: IBM Langflow OSS authentication bypass in MCP composer endpoint
CVE-2026-8446NVD/CVE Database - High
CVE-2026-17630: IBM Langflow OSS remote code execution through configuration parameters
CVE-2026-17630NVD/CVE Database
Exploitation signals
Vulnerabilities published in the week that are listed in the CISA Known Exploited Vulnerabilities catalog or have an EPSS score of 10% or more.No vulnerability published in this week is listed as exploited or has an EPSS score of 10% or more.
Packages that began delegating to a language model
Exposure Registry packages whose first release declaring an LLM SDK, agent framework or MCP dependency was published in the week.| Package | Ecosystem | LLM SDKs | Release | Released |
|---|---|---|---|---|
| vortex-data | PyPI | Hugging Face Hub / Transformers | 0.84.0 |
Topics that moved
Largest increases over the mean of the 4 previous weeks, for topics with at least 3 records in the week.| Topic | Records | Weekly mean, previous 4 | Difference |
|---|---|---|---|
| Frontier model safety | 7 | 2.0 | +5.0 |
| Model Context Protocol | 13 | 9.5 | +3.5 |
| Prompt injection and jailbreaks | 9 | 5.5 | +3.5 |
| Deepfakes and impersonation | 3 | 1.5 | +1.5 |
Research
Peer-reviewed first, then newest.DeepForgeSeal: Latent Space-Driven Semi-Fragile Watermarking for Deepfake Detection Using Adversarial Reinforcement Learning
Peer-reviewedIEEE Xplore (Security & AI Journals)Privacy-Preserving GAN for Synthetic Data against Membership Inference Attack
Peer-reviewedACM Digital Library (TOPS, DTRAP, CSUR)WeatherNext: AI model achieves breakthrough in forecasting cyclones
IndustryDeepMind Safety Research
Policy and regulation
Newest first.AI Therapy under the EU AI Act
EU AI Act UpdatesNCSC statement in response to recent incidents resulting from frontier AI evaluations
UK NCSCWhite House to host AI companies Tuesday to review new model-testing framework
CNBC TechnologyAnthropic, OpenAI among firms facing new scrutiny under EU AI Act enforcement powers
CNBC Technology
Generated from the AI Sec Watch database at . Every item links to its record.