Opinion: lock down MCP servers before an LLM rollout
The piece treats MCP hosts like database drivers and lists controls from input checks to human approval of high-risk actions.
28
Product
Model Context Protocol
- Mentions · 7d3
- Mentions · 30d13
- All mentions33
- First seenAug 11, 2026
- Last mentioned1d ago
Timeline
33 articles · newest firstLog-To-Leak prompt injection extracts MCP agent data after real calls
Across 18 model-server pairs, attack success is 62.64–100%, and defenses still exceed 55% ASR on GitHub MCP with GPT-4o.
64
Matched MCP and CLI interfaces change language-agent behavior, study finds
DART improved scores by up to 9.0 points and cut MCP cost by up to 27.1k tokens per task.
42
Decode-time provenance binding blocks MCP steganographic prompt injection
On a real MCP testbed, 24 adaptive payloads succeeded 100% undefended and 0% with the proposed mechanism.
62
Commentary: AI coding security hinges on adoption maturity, not speed
An eight-stage model from assisted coding to autonomous agents says most developers are still in stages 2 to 4.
22
Stolen AI API keys are paying for hackers’ autonomous scans
Bitsight says criminals sell MONST3R-AI, which uses stolen AI API keys to fund autonomous scans.
64
My Credentials, Please: Detecting Filesystem Capability Laundering in MCP Servers at Scale
Researchers find filesystem capability laundering in MCP servers, including two CVEs in Anthropic's Git server.
66
ServiceNow gives AI agents a passport, and limits on what they can do
ServiceNow launched AI Passport, a runtime credential that limits what enterprise AI agents and MCP servers may do.
40
Researchers bypass 6 AI agent frameworks with a simple ‘Repeat-After-Me’ prompt
Researchers used a Repeat-After-Me prompt to make six AI agent frameworks run attacker-written tool calls.
66
Securing Agent-to-Agent Communication: The Next Identity Frontier
Rapid7 urges identity, least-privilege, and telemetry controls for autonomous AI agent-to-agent communication.
36
Gentlemen Ransomware Affiliate Uses MCP as C2 Channel in Live Cyberattacks
Gentlemen ransomware affiliate Azazel used Model Context Protocol as live command-and-control across victims in six countries.
78
Welcome to the Jungle: What We Found Inside 15,465 Public MCP Servers
OX Security's review of 15,465 public MCP servers found no marketplace vetting and easy domain takeover risks.
55
MCP for agent-to-agent comms may the the riskiest protocol you've never heard of
Researcher finds MCP trust gaps that let prompt injection pivot between agents at Google, Rapid7, and others.
74
MCP Is Creating Major Governance Gaps, Researchers Warn
Ox Security says thousands of MCP servers lack governance and can bypass enterprise security controls.
58
Secrets Sprawl Is an Identity Problem That AI Just Made Impossible to Ignore
GitGuardian finds AI-assisted commits leak secrets about twice as often as human-written code.
58
AI-Powered Threats Exploit Digital Trust Through Autonomous Breach Techniques
A PRC-linked group used Claude Code as an autonomous operator for espionage, while researchers identified the first Android malware leveraging generative AI.
70
A2M: Trace-Optimized Agent Hijacking in the MCP Ecosystem
A2M research shows MCP tool metadata and outputs can hijack agents, with 93.6% malicious invocation on GLM-4.6.
68
Somewhere in your traffic logs, a bot is doing more than looking
Akamai found verified AI crawlers, including ChatGPT, sending frequent POST requests to ecommerce and travel sites.
46
AI Drives Surge in Bot and API Threats
Akamai says AI drove a 300% rise in bot traffic and a 113% increase in daily API attacks.
61
The Target Is No Longer the Model. It’s the Agent.
Analysis maps recent AI security research onto MITRE ATLAS, showing AI agents, poisoned skills and prompt injection form a full attack chain.
55
Vocci’s ring adds a new form factor to meeting note-taking
Vocci launched a $249 titanium AI ring that records and transcribes meetings, raising covert-recording privacy concerns due to its discreet jewelry-like design.
15
Hardcoded MCP credentials found in public GitHub files
Hush Security found 12% of 82,000 public GitHub MCP config files contained hardcoded API keys, bearer tokens, and database passwords.
55
Closed-World Resolution Against Tool Hallucination in LLM Agents
Benchmark across ten LLMs documents 322 tool hallucinations and 154 more on MCP, showing model scale does not help and gates cannot reject fabricated calls.
50
Characterizing Network Centralization and Observability in the Remote MCP Ecosystem
A measurement study of 179 remote MCP servers finds heavy infrastructure concentration (HHI 0.736) and a security-observability tradeoff in platform OAuth.
48
When Agents Look Like Beacons: NIDS Evasion by Model Context Protocol Traffic
Research shows Model Context Protocol agent traffic structurally resembles C2 beaconing and evades Suricata signatures and RITA behavioral scoring in testbeds.
42
OpenAI Builds ‘Defense Factory’ Where AI Agents Continuously Find and Fix Vulnerabilities
OpenAI unveils a Defense Factory where AI agents continuously discover, validate, and fix vulnerabilities, integrating GitHub, Snyk, Semgrep, Tenable, and ServiceNow.
55
iceener/files-stdio-mcp-server: sandbox escape in fs_search via a symlinked directory (recursive walker validates only the top level)
iceener/files-stdio-mcp-server allows a read-side sandbox escape: fs_search follows symlinked directories while the recursive walker validates confinement only at the top level.
35
Off Guard: Breaking LiteLLM from authentication bypass to cloud compromise
Wiz found LiteLLM auth bypass (CVE-2026-59822) and post-auth RCE (CVE-2026-59821) chainable to cloud compromise; the bypass is in CISA KEV with in-the-wild exploitation.
84
Infostealers Target Claude, Cursor, Codex and Other AI Agents to Steal Credentials and Sensitive Data
Gen Digital researchers report infostealer families Amatera, Remus and CallbackBeaver now harvest Claude, Cursor and Codex agent data, including tokens and MCP configs.
62
ToolHive: The open-source way to run any MCP server securely
Stacklok's open-source ToolHive runs Model Context Protocol servers in isolated containers with per-request identity enforcement, audit logging and a signed registry.
45
How MCP Servers Can Expose Enterprise Secrets
MCP servers holding AI agent credentials risk secret exposure via plaintext configs, credential sprawl, prompt injection, and over-permissioning; mitigations include centralization and least privilege.
48
How Cloudflare detects MCP traffic and helps secure it
Cloudflare Gateway now detects Model Context Protocol traffic via protocol-level heuristics to expose shadow MCP usage and enforce approved-server access.
35
Malicious MCP Servers Can Split Instructions to Make AI Coding Agents Exfiltrate Secrets
ASSET Research Group's GhostSplice technique splits malicious instructions across MCP channels, tricking AI coding agents into exfiltrating SSH keys, source code, and secrets.
63
Related CVEs
by mentions- CVE-2026-59822<1%8.8Unauthenticated MCP Endpoint Authentication Bypass in BerriAI LiteLLM
- CVE-2025-541352%9.8Cursor: Cursor allows writing in-workspace files with no user approval in versions below 1.3.9, If the file is a dotfile, editing it…
- CVE-2025-651478%9.6Critical OS Command Injection in mcp-remote Enables RCE via Untrusted MCP Servers
- CVE-2025-327117%7.5Ai command injection in M365 Copilot allows an unauthorized attacker to disclose information over a network
- Rapid7 Bulk Export MCP versions 0.2.5 through 0.6.1 suffer from a GraphQL query injection issue in the export-status component (`get_export_status` in…
- CVE-2026-59821<1%2.1LiteLLM: LiteLLM's Custom Code Guardrails production create and update paths did not apply the same sandboxing and validation used by the…
Appears with
- Anthropicvendor · 6
- Anthropicorganization · 5
- Googlevendor · 4
- OpenAIvendor · 3
- Claude Codeproduct · 3
- DeepSeekai model · 2
- Ox Securityorganization · 2
- Geminiai model · 2
- GPT-4oai model · 2
- GitHuborganization · 2
- Cursorproduct · 2
- Rapid7vendor · 2
- Microsoftvendor · 2
- Rapid7organization · 2
- GitLabproduct · 2
- Akamaiorganization · 2
Entities are extracted by the model from each article. First seen , last mentioned . Watching keeps it in this browser only (no account); the watchlist page and dashboard alerts use it.