--- collection: frontier-ai-labs collection_name: Frontier AI Labs week_start: 2026-08-03 title: Voice, facts, and fixes: the week AI got more careful release_count: 37 generated: 2026-08-10 canonical: https://releases.sh/collections/frontier-ai-labs/digest/2026-08-03 --- # Voice, facts, and fixes: the week AI got more careful OpenAI expanded GPT-5.6 access and made voice conversations far more capable, while Anthropic shipped a thicket of security fixes for Claude Code. The week's story is AI assistants getting both broader and safer. ### A more reliable GPT-5.6, for everyone OpenAI's headline move was pairing a meaningful quality upgrade with wider access. [GPT-5.6 Sol improvements](/release/rel_MJvIDUzUGUQyCknHopyAZ-openai-ships-gpt-5-6-sol-improvements-and-unlimited-gpt-5-6-luna-chats-for-free) landed for Plus and Pro users, promising roughly 68% fewer factual errors on internal evals and more focused answers, plus a new effort slider to dial reasoning depth up or down. It's the kind of incremental but visible polish that makes daily use feel less like gambling on a model's mood. The same change is [documented at the product level](/release/rel_csCORmR29ukl4ZI4OcpXl-chatgpt-updates-gpt-5-6-sol-with-fact-checking-and-reasoning-slider) for anyone tracking the release notes. The bigger structural shift: GPT-5.6 Luna becomes the default for Free and Go users, with unlimited text chats and a "Think" button for harder questions arriving next week — abuse guardrails still apply to file uploads, images, and other tools. That's a genuine expansion of capability to the bottom of the funnel, not just a discount on a paywall. ### Voice gets its hands on files and Projects Across the product surface, voice mode matured from a chat-only channel into something closer to a full workspace. [ChatGPT Voice adds file uploads and Projects support](/release/rel_IbtCr9E0DRXLVuYaHhUHf-chatgpt-voice-adds-file-uploads-and-projects-support), letting you upload images, docs, and PDFs mid-conversation and analyze them aloud, or drop into a Project and reference its recent chats, sources, and instructions. The same capability landed for [ChatGPT Business](/release/rel_rICHdkQQ6kWimgRbW8_tm-chatgpt-business-adds-file-uploads-and-projects-support-to-voice) and for [Enterprise, Edu, and Healthcare workspaces](/release/rel_Z5QVTMtfa5vazVK9Ax5bk-chatgpt-enterprise-edu-live-becomes-default-voice-experience-files-and-projects), where "Live" also becomes the default Voice experience when admins enable it — no more Early Model Access toggle. Voice is now a first-class way to work, not a parlor trick. Two other productivity notes: the [ChatGPT for PowerPoint add-in reached general availability](/release/rel_qVlxAhY030b8RI43_rV_G-chatgpt-for-powerpoint-add-in-reaches-general-availability), letting you build and revise decks from a sidebar while keeping slides editable in PowerPoint itself — a meaningful step toward agentic work inside the Microsoft Office workflow. And [Sign in with ChatGPT](/release/rel_YFnTmMfBMo2x-KYjGbYYm-chatgpt-launches-sign-in-with-chatgpt-across-partner-sites) began rolling out across Airtable, GitLab, HubSpot, Notion, Supabase, and Vercel, which is OpenAI quietly building an identity layer for the connective tissue of the modern workplace. ### Claude Code hardens its sandbox While OpenAI pushed capability forward, Anthropic spent the week making Claude Code harder to break out of. The back-to-back releases read like a security audit in real time. [Claude Code v2.1.221](/release/rel_Z0WSuG0O0Qh38qf7AjknN-claude-code-v2-1-221-fixes-bash-tool-permission-bypass-adds-focus-view-and) fixed a Bash tool permission-check bypass where zsh could slip hidden commands through regex conditionals, added a Focus view to collapse tool activity, and introduced sandbox credential file masking on Linux and WSL. The very next day, [v2.1.222](/release/rel_BFpHVqOccghRUA9vLslSx-claude-code-v2-1-222-fixes-worktree-isolation-and-hook-bypass-vulnerabilities) closed a worktree-isolation gap where destructive git commands could hit the main checkout, plus a PreToolUse auto-allow hook bypass in background tasks. Then [v2.1.223](/release/rel_G1hDm6xoiapkV3NoQGHrg-claude-code-v2-1-223-fixes-bash-permission-bypass-and-sandbox-escapes) fixed another Bash permission bypass and pulled 1M-context models back to a 200K auto-compaction window, while [v2.1.224](/release/rel_PGWWietOyhv80YsFey6HI-claude-code-v2-1-224-adds-self-hosted-environments-and-cross-session-messaging) added self-hosted environments via `claude self-hosted-runner` and cross-session messaging with SendMessage. The pattern — bypass, fix, new bypass, fix — is exactly what you'd expect from a tool that's becoming a trusted agent running arbitrary commands on developer machines. The [v2.1.225 OAuth token loss fix](/release/rel_DNbLbsOPB518lQ0-NVjiO-claude-code-v2-1-225-fixes-oauth-token-loss-and-mcp-401-bursts) and the reliability-focused [v2.1.226](/release/rel_Fka9ZAKrEADUtZJx-Wt9D-claude-code-v2-1-226-reliability-improvements) round out the week's churn. Security patches aren't glamorous, but for anyone running Claude Code unattended, they're the most important releases here. ### Supporting movement across the rest of the stack The lower-key releases still tell a coherent story of incremental polish. [o1 and o3-mini gained Python-powered data analysis](/release/rel_AA5LRa11TIieY3y2RyrAm-chatgpt-o1-and-o3-mini-gain-python-powered-data-analysis) in ChatGPT, [Codex CLI 0.147.0](/release/rel_MqUbl76bzPq58lF-1glNT-codex-cli-release-0-147-0) added portable Agent Plugins and persistent conversation sections, and the [Node SDK v7.4.0](/release/rel_jUUez9zGYl7fSrdcASfN--openai-node-sdk-v7-4-0-adds-gpt-5-5-model-and-tool-metadata-fields), [Python SDK v2.53.0](/release/rel_g4QybgMZukof5GGrYQg4q-openai-python-sdk-v2-53-0-adds-gpt-5-5-and-tool-name-namespace-to-responses), and [Go SDK v3.50.0](/release/rel_RpVAkos96XYmvWrKO_8A4-openai-go-sdk-v3-50-0-api-changes) all picked up gpt-5.5 support and tool metadata fields — the plumbing for the model updates above. [Enterprise & Edu admin controls](/release/rel_51PVnefE9QHpMs_d_mjrI-chatgpt-enterprise-edu-admins-can-now-disable-desktop-app-auto-updates) for desktop app updates and [education plugins](/release/rel_6VDz3PjfvpzTPCIeeGZqu-chatgpt-enterprise-edu-launches-education-plugins-for-teaching-and-learning) for teachers and students round out a week where the big players expanded both reach and guardrails. ## Releases covered ### Anthropic - [Claude Code v2.1.221 fixes Bash tool permission bypass, adds Focus view and sandbox credential masking](https://releases.sh/release/rel_Z0WSuG0O0Qh38qf7AjknN-bash-tool-permission-bypass-fixed-sandbox-credential-masking-shipped) - [Claude Code v2.1.222 fixes worktree isolation and hook bypass vulnerabilities](https://releases.sh/release/rel_BFpHVqOccghRUA9vLslSx-worktree-isolation-strengthened-hook-bypass-fixed) - [Claude Code v2.1.223 fixes Bash permission bypass and sandbox escapes](https://releases.sh/release/rel_G1hDm6xoiapkV3NoQGHrg-bash-permission-bypass-fixed-sandbox-escapes-closed) - [Claude Code v2.1.224 adds self-hosted environments and cross-session messaging](https://releases.sh/release/rel_PGWWietOyhv80YsFey6HI-self-hosted-runners-and-cross-session-messaging-arrive) - [Claude Code v2.1.225 fixes OAuth token loss and MCP 401 bursts](https://releases.sh/release/rel_DNbLbsOPB518lQ0-NVjiO-oauth-token-loss-fixed-mcp-401-bursts-resolved) - [Claude Code v2.1.226 reliability improvements](https://releases.sh/release/rel_Fka9ZAKrEADUtZJx-Wt9D-reliability-improvements) ### OpenAI - [OpenAI ships GPT‑5.6 Sol improvements and unlimited GPT‑5.6 Luna chats for free users](https://releases.sh/release/rel_MJvIDUzUGUQyCknHopyAZ-gpt-5-6-sol-more-factual-and-focused-luna-unlimited-for-free-users) - [ChatGPT updates GPT‑5.6 Sol with fact-checking and reasoning slider](https://releases.sh/release/rel_csCORmR29ukl4ZI4OcpXl-gpt-5-6-sol-more-reliable-focused-answers-reasoning-slider-new) - [ChatGPT Voice adds file uploads and Projects support](https://releases.sh/release/rel_IbtCr9E0DRXLVuYaHhUHf-voice-chats-gain-file-uploads-and-project-awareness) - [ChatGPT Business adds file uploads and Projects support to voice](https://releases.sh/release/rel_rICHdkQQ6kWimgRbW8_tm-voice-now-supports-file-uploads-and-projects) - [ChatGPT Enterprise & Edu: Live becomes default Voice experience, files and Projects arrive in voice](https://releases.sh/release/rel_Z5QVTMtfa5vazVK9Ax5bk-voice-defaults-to-live-files-and-projects-supported) - [ChatGPT for PowerPoint add-in reaches general availability](https://releases.sh/release/rel_qVlxAhY030b8RI43_rV_G-chatgpt-for-powerpoint-now-ga) - [ChatGPT launches Sign in with ChatGPT across partner sites](https://releases.sh/release/rel_YFnTmMfBMo2x-KYjGbYYm-sign-in-with-chatgpt-rolls-out-to-six-partner-sites) - [ChatGPT o1 and o3-mini gain Python-powered data analysis](https://releases.sh/release/rel_AA5LRa11TIieY3y2RyrAm-o1-and-o3-mini-now-run-python-data-analysis) - [Codex CLI Release: 0.147.0](https://releases.sh/release/rel_MqUbl76bzPq58lF-1glNT-codex-cli-release-0-147-0) - [OpenAI Node SDK v7.4.0 adds gpt-5.5 model and tool metadata fields](https://releases.sh/release/rel_jUUez9zGYl7fSrdcASfN--gpt-5-5-model-and-tool-metadata-fields-added) - [OpenAI Python SDK v2.53.0 adds gpt-5.5 and tool name/namespace to Responses](https://releases.sh/release/rel_g4QybgMZukof5GGrYQg4q-gpt-5-5-model-added-tool-name-namespace-fields-in-responses) - [OpenAI Go SDK v3.50.0 API changes](https://releases.sh/release/rel_RpVAkos96XYmvWrKO_8A4-go-sdk-api-changes-promoted) - [ChatGPT Enterprise & Edu admins can now disable desktop app auto-updates](https://releases.sh/release/rel_51PVnefE9QHpMs_d_mjrI-desktop-app-auto-updates-now-toggleable-by-admins) - [ChatGPT Enterprise & Edu launches education plugins for teaching and learning](https://releases.sh/release/rel_6VDz3PjfvpzTPCIeeGZqu-college-educator-k-12-educator-and-college-student-plugins-launch)