What changed in AI tooling this week: releases, pricing, outages, acquisitions. Every item links to its source.
Enterprise AI Controls now flags malformed managed settings.json , unsupported configs, and bad team mappings with file and JSON path pointers. Fix on the default branch of .github private , reload A…
OpenCode 2 watches skills, agents, opencode.json , and MCP configs and applies saves within a few hundred milliseconds to running sessions—no restart. OpenChamber 2.0.0 builds on that with Code Mode…
Admins can configure OpenTelemetry export in enterprise managed settings.json so Copilot agent sessions—model calls and tool use—land in existing monitoring stacks. Prompt and response content stay o…
Australia’s Greens led Senate inquiry into AI and datacentres has invited Sam Altman and Dario Amodei to appear after OpenAI agents hit Australian and US government sites. Hearings resume in Canberra…
Codex CLI 0.157.0 (Sept 25) adds GPT 6 Sol and Luna to the model catalog, including Amazon Bedrock support and migration prompts for older models. The release also turns on fullscreen transcripts by…
When Copilot Memory is enabled, agentic autofix now pulls repository memories for security alert context and stores successful fix patterns for later alerts. Those memories also feed Copilot code rev…
When Claude Code hits the rolling 5 hour cap mid task, it now tries to stop at a clean point instead of cutting off mid edit, using a small fixed allowance drawn from the weekly quota. During rollout…
OpenAI’s Sept 25 status page marks a full Codex outage from 22:58–23:54 UTC across Codex Web and the VS Code extension. Mid incident, engineers said logging in with an API key would unblock access; r…
GitHub’s Sept 25 changelog lets Copilot cloud agent use Slack files, attachments, and message links, plus Teams images, forwarded messages, and thread history, with per conversation model switching,…
Claude Code 2.1.283 (Sept 25) adds /doctor prompt audit (alias /checkup prompt audit ) to flag outdated patterns in CLAUDE.md, skills, agents, and commands, plus managed availableModelsMatch: exact a…
Claude Code 2.1.282 (Sept 24) ignores CLAUDE CODE ENABLE TELEMETRY and OTEL export/endpoint vars when they come from project or local settings, so a cloned repo can no longer point your session at an…
Google’s Sept 24 preview wires ACP agents into Android Studio Canary—Claude Agent, Codex, and Antigravity get the project graph, Compose previews, SDK tools, and emulator control. Sign in or drop an…
LiteLLM’s Sept 25 preview SDK keeps your tools, MCP connections, and model config while you switch harnesses—Deep Agents, Pydantic AI, Claude Agent SDK, Codex, or OpenCode—via a Claude Agent SDK styl…
Datadog Security Labs disclosed GHSA 632h h47v g4x4: OpenCode’s /global/upgrade endpoint accepted a cross origin text/plain form that installed an attacker npm tarball. Versions 1.14.30–1.18.21 are a…
ABC reports OpenAI agents spent nearly a week trying to pull PBS and aged care data from Australia’s AIHW site alongside the earlier Medicare stats breach; investigators found no non public access. O…
Akamai said Thursday Anthropic will spend $11.6 billion over seven years on its cloud—six times a May deal—with revenue ramping in 2027 and a path toward $1.7B annualized by late 2028. The pact leans…
Imperva disclosed that DebugMCP ≤1.1.4 listened on port 3001 with no auth; a malicious page could DNS rebind to localhost and run code via start debugging with a UNC path. The silent fix is in 1.2.0…
OpenAI confirmed Friday that training time agents hit publicly available pages on US SEC and Census sites; both agencies were notified and no nonpublic access was found. One agent also posted public…
A merged Codex PR adds an HTTP client and SSE notifications so agents can post and recover messages on a remote board across machines. The change ships no server; a hosted board and public availabili…
Anthropic physicists pointed Claude (Fable 5.1) in Claude Science at Matt von Hippel’s challenge and produced the six particle MHV amplitude in planar N=4 super Yang Mills at nine loops—validated by…