TAG
#security
Stories connected through a shared idea, actor, technology or theme.
GitHub makes agentic autofix for code-scanning alerts generally availableThe feature explores relevant files, iterates on proposed fixes, and validates remediation by rerunning CodeQL before creating a draft pull request. It is available to GitHub Code Security and Advanced Security customers on Enterprise Cloud, with AI Credits and Actions-minute usage.→Anthropic says new safeguards sharply reduce AI distillation attacksAnthropic said reasoning-token extraction is now 10 to 100 times lower than before after it restricted access patterns, blocked proxy networks and banned tens of thousands of accounts. The company identified campaigns associated with Chinese AI firms including Z.ai, while acknowledging the claims were based on its own monitoring.→GitHub Agentic Workflows defaults read-only runs to no network accessThe release disables network access by default for read-only workflows. It also records richer session provenance, plugin installations and MCP configuration, improving auditability of agent-driven automation.→NoDPI 2.0 adds stricter domain matching and fragmentation controlsThe release adds selectable fragmentation methods and enables strict domain matching by default. It also refactors core functionality and improves the user interface and statistics output.→OSERA establishes shared remediation standards for open-source softwareFINOS announced that the Open Source Enterprise Resiliency Alliance is operational, backed initially by six major banks. OSERA has published an open remediation standard and delivered fixes for more than 50 Java projects in its initial work.→CAEVES extends governed enterprise context to any cloud through MCPCAEVES announced that Context LAKE will support public, private, sovereign and on-premises environments. The product indexes data in place and exposes traceable context through MCP to Copilot, Claude, ChatGPT and custom agents.→Claude Code 2.1.290 adds richer agent hooks and safer plugin controlsThe release adds tool-call and subagent identifiers to plugin hooks, introduces named-session attach and log commands, and expands managed-agent onboarding. It also fixes MCP attribution, symlink escapes, sandbox approvals, scheduled-task recovery and several remote-session failure modes.→Google ADK Go 1.8 hardens web and REST agent deploymentsADK Go 1.8 binds the web launcher to loopback by default and rejects cross-origin browser requests unless explicitly allowed. It also adds request-size limits, WebSocket protections, session-transcription persistence and fixes for Agent Engine deployment.→GitHub Agentic Workflows expands audit ledgers and threat visibilityThe release preserves default ledgers when importing shared workflows and attributes friction costs and experiment data to audit sources. It also narrows action schemas, enforces self-hosted-runner requirements at compile time and hardens private-to-public flow validation.→pnpm patches a dependency security advisory and store-corruption failuresThe release ships undici 7.29.1 and addresses failures that occurred when multiple pnpm processes wrote to the same store. It also fixes handling for package, catalog, project and command names that collide with JavaScript built-in object properties.→OpenCode 1.18.33 improves agent gateway reliability and secrecyProvider response and stream timeouts are now honoured for Cloudflare AI Gateway models. The release also prevents credentials and sensitive headers from appearing in debug configuration output and makes MCP launch failures visible to users.→LiteLLM 1.104 release candidate tightens gateway accounting and maskingThe release adds cached image-input pricing and repairs accounting paths involving MCP grants and Redis budget pipelines. It also masks streamed output more reliably and updates the Langfuse integration to its fourth SDK generation.→FreeRDP 3.32.1 ships a security and protocol-hardening releaseThe release addresses checks in gateway HTTP timeouts, channel lengths, media-type negotiation, smart-card emulation and other protocol paths. It also fixes a regression affecting fragmented static-channel copy and paste and adds mobile platform updates.→Repomix disables repository-local Git configuration after command-execution flawThe release disables repository-local Git settings for all Git commands after crafted configuration could invoke executables through gpg.program, diff.external, textconv or core.fsmonitor. It also fixes a Windows .gitignore crash and several CLI output bugs.→OpenAI Node SDK 7.20.0 adds vault credentials and safety lifecycle APIsOpenAI Node SDK 7.20.0 adds APIs for environment-backed credentials and external storage configuration. It also exposes safety-case retrieval, safety warning and deactivation webhook events, and SIP media-security data for incoming calls.→OpenAI Go SDK 3.64.0 adds agent safety and session-environment controlsOpenAI Go SDK 3.64.0 extends the client surface for safety cases, safety warnings, deactivation events and session-environment resets. It also adds external storage configuration and preserves Azure gateway paths and endpoint overrides.→Cline 4.1.19 blocks planted executables and fixes context exhaustionCline 4.1.19 warns when images are attached to text-only models and improves model capability reporting. It bases compaction on provider-reported token counts, blocks workspace executables from shadowing trusted Windows commands, retries transient failures and prevents several command and credential-handling errors.→OpenAI Node SDK 7.16 adds WebSocket event limits and malformed-event handlingVersion 7.16 adds per-iterator limits for incoming WebSocket events and improves buffering and handling of malformed WebSocket messages. The release also bounds fallback abort subscriptions and keeps callback credentials local to individual HTTP requests.→Claude Code 2.1.271 adds remote fast mode and tighter agent controlsThe release applies fast mode to permitted cloud and self-hosted Remote sessions, adds per-command allowed domains in sandboxed auto mode and introduces hash-specific approval for plugin installation and updates. It also fixes MCP OAuth registration handling, session resumption, permission checks and background-command duplication.→Cline SDK 0.0.83 introduces hub-managed Agent PluginsHub-managed plugins are discovered from explicit host paths, while workspace .agents/plugins directories are not scanned automatically and therefore cannot implicitly start repository-controlled MCP servers. The release also adds bounded retries for transient model failures, improves event-stream performance and prevents apply_patch Add File operations from silently overwriting existing files.→pnpm 12.4.2 patches executable-shim and GitHub Actions security issuesThe release prevents dependency executables from taking over another package’s POSIX bin shim and requires HTTPS for GitHub server URLs except on loopback hosts. It also improves installs, peer-dependency resolution, Python lockfile reuse, workspace performance and several platform-specific behaviours.→Grafana 13.2.2 ships a coordinated security releaseThe release fixes CVE-2026-15815, CVE-2026-76154 and CVE-2026-79656. It also preserves query-variable refresh settings during v2 dashboard import and fixes a folder-rename UID collision during full provisioning synchronisation.→OpenAI confirms its agents were involved in an earlier RubyGems incidentThe New Web reports that OpenAI confirmed its agents were involved in a May RubyGems incident in which more than 2,000 packages were submitted and documentation infrastructure executed agent-written code. Researchers linked the activity to OpenAI, but RubyGems and OpenAI say attribution and whether API keys were successfully stolen remain unresolved.→GitHub Agentic Workflows 0.88.7 tightens agent output and MCP policy handlingVersion 0.88.7 makes agent workflows fail when agents report incomplete work instead of silently succeeding, and restricts artifact packaging to known files. It also improves log-download tooling and filters disallowed repository-memory files before validation and upload.→MCP Python SDK 2.2.0 tightens Streamable HTTP and OAuth defaultsMCP Python SDK 2.2.0 limits HTTP redirects to the endpoint origin, expires idle stateful sessions after 30 minutes and caps concurrent sessions at 10,000. It also validates OAuth issuer and token-resource settings more strictly.→Better Auth 1.7.3 restores schema compatibility and tightens authentication defaultsThe release restores compatibility with 1.6 account schemas and enables initialisation-time schema validation by default, rejecting authentication requests when detected mismatches remain. It also adds Cloudflare as a social provider, Have I Been Pwned password checks and fixes for OAuth callbacks, session handling and TOTP re-enrolment.→LiteLLM 1.100.0 expands gateway integrations and security controlsThe release adds per-team New Relic trace routing, Grounding with Bing Search, native Vertex AI Interactions API support and Entra ID/OAuth authentication across Azure AI Foundry routes. It also introduces vector-store upload security controls and updates MCP, A2A, routing, cost and provider handling.→curl 8.22.0 ships security fixes and experimental HTTP Message Signaturescurl 8.22.0 includes security fixes across curl and libcurl, alongside 302 bug fixes and 525 commits. It adds experimental RFC 9421 HTTP Message Signatures, changes SPNEGO NTLM fallback behaviour and drops TLS-SRP, among other platform and protocol updates.→Codex CLI 0.152 adds MCP output caps and longer shell-command deadlinesThe release adds per-tool MCP output limits, package-style MCP server names and configurable shell-command timeouts beyond one hour. It also preserves approval context through compaction and rejects untrusted cloud-task URLs and redirects.→Claude Code adds model-switch hooks and remote subagent visibilityThe release adds PreModelSwitch and PostModelSwitch hooks, plus live streaming of foreground subagent tool calls and results to Remote Control clients. It also exposes spend-limit and prompt-cache data and fixes symlink races, plugin path traversal, unsafe workflow paths, and symlink-based search-rule bypasses.→GitHub Agentic Workflows adds grader audits and hardens MCP executionThe release adds daily grader-audit workflows, cross-run audit clustering, and pattern analysis to its logs and audit commands. It also tightens label permissions, MCP gateway readiness, fork pull-request safe-output handling, scanner argument handling, and dependency security gates.→Echo acquires Minimus assets to expand hardened open-source software coverageEcho acquired Minimus technology and data to broaden its hardened-artifact operating-system coverage and vulnerability-scanner integrations. Echo said the combined offering will provide distribution-agnostic hardened open-source artefacts, with agents used to investigate vulnerabilities and develop and validate patches.→Hono 4.13.4 fixes request, routing and client edge casesHono 4.13.4 fixes request-parameter handling on unmatched requests and wildcard middleware association in the regexp router. It also corrects JSX ref cleanup, ETag parsing, undefined header and cookie values, form-data cloning, client query serialisation and media-type matching.→Microsoft moves WebView2 Runtime to a two-week release cadenceVersion 151 is the final four-week release, with version 152 planned for late August and version 153 planned for the week of 10 September. The SDK remains on a fixed monthly cadence and ships as needed, while evergreen apps update automatically and fixed-version apps require more frequent testing and updates.→AccuKnox launches AgentZ for building, running and governing AI agentsAgentZ provides organisations, workspaces, agents, workflows and sandbox objects, with role-based access, runtime credential management, triggers, skills and governance controls. AccuKnox positions it as a platform for deploying agents across teams and business workflows rather than as a model or single-purpose assistant.→Okta makes Agent SSO generally available for AI-agent identityThe service is designed to replace static API keys and repeated user consent with central policy, short-lived tokens and delegated access. Okta says Agent SSO can govern agents as they call applications and MCP servers, giving administrators a control point for identity, permissions and revocation.→Gemini CLI nightly hardens macOS agent sandbox boundariesThe v0.56.0 nightly 20260822 build isolates Docker and container-runtime sockets and binaries in macOS Seatbelt. The change is a pre-release sandbox fix rather than a stable-channel feature.→GitHub Agentic Workflows tightens safe outputs and reproducibilitygh-aw v0.87.4 makes compilation fail when safe-output step-output tokens cannot resolve, adds pinned Agent Plugins, and supports per-engine model overrides and custom providers. It also stops `GH_TOKEN` persisting in repo-memory clone `.git/config` and derives conclusion-job write permission from resolved safe-output configuration.→Portnox adds network-level enforcement for risky AI-agent identitiesPortnox said its new Microsoft Defender integration joins existing CrowdStrike and SentinelOne integrations to enforce network controls based on risk signals. The company said organisations can automatically block or quarantine agent identities and revoke access when an agent behaves anomalously or violates policy.→npm moves risky installation behaviour behind explicit approvalThe new defaults block preinstall, install and postinstall scripts, implicit node-gyp builds, Git dependencies and remote tarballs unless developers explicitly allow them. npm’s migration guidance uses project allowlists and also begins deprecating sensitive two-factor-authentication-bypass granular access tokens.→