Dateline: May 29, 2026 | Next update: June 5, 2026
The headline this week is a new model: Claude Opus 4.8 launched on May 28, becoming the most capable generally available Claude with a 4x improvement in catching its own code errors and a new Dynamic Workflows feature for large-scale parallel agent orchestration. Alongside the model launch, Project Glasswing published its first major progress report — Claude Mythos Preview found over 10,000 high- or critical-severity vulnerabilities in critical open-source software in one month — and Claude Compliance API integrations arrived for enterprise IT and security governance. Claude Code also shipped a real-time security plugin and a broad stability update.
Claude / Anthropic
Claude Opus 4.8 — new flagship model
Claude Opus 4.8 is now the most capable generally available Claude model, shipping just weeks after Opus 4.7. The headline improvement is honesty about its own work: Anthropic reports it is four times less likely than Opus 4.7 to let flaws in code it has produced go unremarked. It also ships three new API capabilities — mid-conversation system messages, public stop_details documentation, and Dynamic Workflows — and arrives at the same price as Opus 4.7, with a significantly cheaper fast mode.
Opus 4.8 launches May 28 with the following key changes over Opus 4.7: (1) 4x less likely to silently overlook defects in its own code. (2) Dynamic Workflows (research preview, Claude Code only): ask Claude to create a workflow and it orchestrates tens to hundreds of parallel subagents in the background, using existing test suites as success signals — target use case is codebase-scale migrations across hundreds of thousands of lines. (3) Mid-conversation system messages: you can now send role: "system" messages after a user turn in the messages array, preserving prompt cache compatibility. (4) stop_details now publicly documented on refusal responses — applications can route refusals to the correct next-step UX without parsing string heuristics. Opus 4.8 fast mode arrives at $10/$50 per MTok, 3x cheaper than previous Opus fast modes.
Model string: claude-opus-4-8 | Pricing: $5/$25 per MTok (standard) | Fast mode: $10/$50 per MTok | Context: 1M tokens (API, Bedrock, Vertex AI), 200k (Microsoft Foundry) | Max output: 128k tokens | Mid-conversation system messages: role: "system" after user turn, placement rules apply | stop_details: documented refusal categories, no beta header required | Dynamic Workflows: research preview, Claude Code only | CLAUDE_CODE_OPUS_4_6_FAST_MODE_OVERRIDE deprecated, removal June 1
Best for: Developers needing the highest-quality code review and self-verification; teams running large-scale automated migrations with Dynamic Workflows
Project Glasswing — 10,000+ critical vulnerabilities found
Anthropic published its first detailed Project Glasswing progress report, revealing that roughly 50 partner organisations — including AWS, Apple, Broadcom, Cisco, Google, JPMorgan Chase, the Linux Foundation, Microsoft, NVIDIA, CrowdStrike, Cloudflare, and Mozilla — have used Claude Mythos Preview to find over 10,000 high- or critical-severity software vulnerabilities in one month.
23,019 total vulnerabilities identified across 1,000+ open-source projects; 6,202 classified high or critical. Of 1,752 independently assessed by six security research firms, 90.6% (1,587) are confirmed true positives; 62.4% (1,094) confirmed high or critical severity. Anthropic projects roughly 3,900 confirmed high/critical issues in open source at current rates. Standout results: Cloudflare found 2,000 vulnerabilities (400 high/critical) with a false-positive rate better than human testers; Mozilla found 271 vulnerabilities in Firefox 150 — more than 10x the number found in Firefox 148 with Claude Opus 4.6. One banking partner used Mythos to detect and prevent a fraudulent $1.5M wire transfer. A critical wolfSSL flaw (CVE-2026-5194) was discovered in a library embedded in approximately 5 billion IoT and industrial devices. Anthropic is now partnering with the Open Source Security Foundation's Alpha-Omega project to help open-source maintainers process the patching backlog.
Glasswing model: Claude Mythos Preview (not publicly available) | Partners: ~50 organisations including AWS, Apple, Cisco, Google, Microsoft, NVIDIA, CrowdStrike, Cloudflare, Mozilla | True-positive rate: 90.6% of assessed findings | Disclosure policy: standard 90-day coordinated disclosure; patched vulnerabilities disclosed ~45 days post-patch | anthropic.com/glasswing now live | Anthropic has revised its position to allow partners to share findings with regulators and the public
Best for: Security teams, infrastructure owners, and anyone running software built on audited open-source libraries — expect an unusually heavy patch season in H2 2026
Claude Compliance API integrations — enterprise governance
Anthropic launched Claude Compliance API integrations this week, allowing enterprise IT and security teams to govern Claude across the full Claude platform and product suite using the same tools they already use to govern other applications in their stack.
IT and security teams can now apply existing security and compliance tooling to Claude across the platform — covering policy enforcement, audit logging, and access controls. Integration guides available in Claude Docs under Claude Compliance API integrations.
Platform: Claude Platform (all products) | Access: Enterprise | Documentation: Claude Compliance API integrations in Claude Docs | Complements: existing Claude Console audit logs, CloudTrail on Claude Platform on AWS, SCIM/RBAC in Cowork
Best for: Enterprise IT and security teams, compliance officers, regulated industries (finance, healthcare, legal, government)
Claude Code — real-time security plugin
A new security-guidance plugin for Claude Code brings real-time vulnerability detection directly into the coding session. It monitors code edits, diffs, and commits as they happen and flags dangerous patterns before they reach production.
Security-guidance plugin now available in the Claude Code plugin marketplace — install with /plugins. Detects approximately 25 high-risk vulnerability classes in real time including SQL injection, command injection, cross-site scripting, hardcoded API keys and secrets, insecure deserialisation, and improper input validation. When a risky pattern is detected, Claude prompts an inline fix in the same session. Internal data shows a 30–40% reduction in security-related PR comments since the tool was introduced. Enable the new /tui feedback toggle — Anthropic intends to make it default in Claude Code soon.
Install: /plugins in Claude Code | Detection method: regex-based pattern matching across ~25 vulnerability classes | Inline fix: prompted within the same session | Complements: Claude Security (enterprise, deeper AI-driven codebase scanning with data-flow tracing and adversarial verification) | /tui feedback: new toggle, planned to become default
Best for: All developers using Claude Code who want real-time security feedback without switching to a separate scanner
Claude Code — usage insights, diff navigation, and security hardening
Claude Code shipped a broad point release adding richer usage visibility, keyboard-driven diff navigation, improved Markdown rendering, and several security and stability fixes.
/usage now shows a per-category breakdown of what is driving your limits — skills, subagents, plugins, and per-MCP-server cost broken out separately. /diff detail view can now be scrolled with the keyboard (arrows, j/k, PgUp/PgDn, Space, Home/End). Markdown output now renders GFM task-list checkboxes (- [ ] / - [x]) as checkboxes. Enterprise setting added for Claude.ai cloud MCP connectors. Auto-mode classifier improved for detecting data exfiltration, particularly bulk transfers of repository contents. Security fix: rm -rf $HOME no longer bypassed when HOME has a trailing slash. Security fix: TMPDIR no longer resolves to different directories in sandboxed vs unsandboxed Bash commands within the same session. CLAUDE_CODE_SUBAGENT_MODEL now correctly applies to teammate processes spawned by Agent Teams.
/usage: per-category breakdown (skills, subagents, plugins, per-MCP-server) | /diff: keyboard navigation added | GFM task lists: now rendered as checkboxes | Fixed: CLAUDE_CODE_SUBAGENT_MODEL not applying to Agent Teams teammate processes | Fixed: slash commands followed by tab or newline treated as unknown command | Fixed: rm -rf $HOME trailing-slash bypass | Fixed: TMPDIR inconsistency across sandboxed/unsandboxed Bash | Fixed: spacing/layout glitches in /plugin, /status, /mobile, /sandbox, /permissions menus
Best for: All Claude Code users, especially those monitoring usage limits, running Agent Teams, or working in sandboxed environments
Anthropic surpasses OpenAI in US enterprise AI adoption
A third-party enterprise AI adoption survey published this week shows Anthropic has overtaken OpenAI in US enterprise AI adoption, with a 34.4% share versus OpenAI's 32.3%. This is the first time Anthropic has led this metric.
Anthropic now leads US enterprise AI adoption: 34.4% share vs OpenAI at 32.3%, per third-party survey data. Claude Code is cited as the primary driver among engineering teams. Contributing factors: Claude Platform on AWS (GA May 11), PwC alliance (May 14), SpaceX compute deal (May 6).
Source: third-party enterprise adoption survey (not Anthropic self-reported) | Claude Code run-rate: approximately $2.5B annualised (Feb 2026) | Anthropic overall run-rate: $30B+ (Apr 2026)
Best for: Enterprise buyers, competitive intelligence, investors
Plans and Pricing
Opus 4.8 launches at the same standard price as Opus 4.7 ($5/$25 per MTok). The fast mode is meaningfully cheaper: $10/$50 per MTok — 3x cheaper than the previous Opus fast mode rate. No other pricing changes this week.
Opus 4.8 standard: $5/$25 per MTok | Opus 4.8 fast mode: $10/$50 per MTok (3x cheaper than prior Opus fast mode) | Sonnet 4.6: $3/$15 per MTok | Haiku 4.5: low-cost tier | Opus 4.6 fast mode override deprecated June 1 — switch to /model claude-opus-4-6[1m] then /fast
ChatGPT / OpenAI
Dateline: May 29, 2026 | Next update: June 5, 2026
Over the past week, OpenAI has focused on expanding ChatGPT's role as a workplace platform, improving enterprise collaboration features, and continuing to refine agentic workflows.
GPT-5.3 Standard — default model
Improved consistency across mixed workflows involving documents, images, and web-based tasks.
Better handling of context switching within long conversations.
Context ~128k | Output ~4k–8k | Improved context retention mechanisms
Best for: General use
GPT-5.3 Pro — high-reasoning model
More reliable performance on long-form analytical and strategic work.
Improved reasoning stability during extended multi-step tasks.
Context ~200k (est.) | Better long-horizon reasoning consistency
Best for: Deep analysis
GPT-5.3 Mini — fallback model
Improved routing efficiency, reducing unnecessary model switching.
Context ~64k | More efficient inference allocation
Best for: Quick tasks
Agent Mode
Improved ability to manage multi-stage projects involving research, drafting, and revision.
Enhanced workflow persistence | Better task-state tracking
Best for: Task delegation
Deep Research
Improved source synthesis when analysing large volumes of information.
Enhanced ranking and consolidation pipeline | Better conflict resolution between sources
Best for: Research
Memory & Projects
Improved project-level memory management and organisation.
Better separation of project context from global memory | Improved retrieval relevance
Best for: Ongoing workflows
Advanced Voice Mode
More natural handling of interruptions and follow-up questions.
Improved conversational state tracking | Reduced latency during turn-taking
Best for: Voice interaction
ChatGPT for Work
OpenAI continues positioning ChatGPT as a collaborative workplace platform.
Expanded emphasis on team collaboration, shared workflows, and project-centric workspaces.
Improved project architecture | Enhanced collaboration features across teams
Best for: Organisational productivity
Plans and Pricing
No significant pricing changes announced this week. Pricing stable | API structure unchanged.
Gemini (Google)
Date: May 29, 2026 | Next update: June 5, 2026
Following the massive wave of announcements at Google I/O, this week focused on the stabilisation of the Gemini 3.5 infrastructure, early developer feedback on the Antigravity suite, and the conclusion of several major API transitions.
Gemini 3.5 Flash — production stabilisation
Infrastructure has successfully scaled to handle the post-I/O traffic surge. Early developer benchmarks confirm the model's targeted four-fold speed increase in output token generation.
Production infrastructure successfully scaled following post-I/O traffic spike. Early developer benchmarks confirm 4x output token speed improvement. Standard API endpoint uptime at steady 99.9%.
Best for: High-volume production workloads
Gemini 3.1 Flash-Lite — final shutdown
As scheduled, the legacy gemini-3.1-flash-lite-preview model was permanently turned off on May 25, 2026. All production traffic must now point to the General Availability (GA) version.
gemini-3.1-flash-lite-preview permanently turned off May 25, 2026. All traffic must migrate to the GA model immediately.
Best for: Immediate action required if still using the preview model
Gemini 3.5 Pro — testing continues
Internal tuning continues for the upcoming Pro model, focusing on expanding the multi-tab "Skills in Chrome" capabilities before its widespread release next month.
Best for: Watch this space — no action needed yet
API schema shift — breaking change
A major, breaking update to the Interactions API went live on May 26, 2026. The response schema has officially migrated from outputs to steps to better accommodate multi-stage agent workflows.
Interactions API response schema migrated from outputs to steps. Breaking change — update all code that reads the outputs field. Required to support multi-stage agent workflows.
Best for: Agent developers — immediate code update required
Antigravity CLI — rapid developer adoption
Following the rollout of the Antigravity 2.0 developer suite at Google I/O, Google reported a rapid transition of active developers moving workflows over from the legacy Gemini CLI. All Gemini CLI users are urged to migrate immediately.
Rapid developer migration from Gemini CLI to Antigravity CLI underway. Legacy Gemini CLI workflows should be migrated now.
Best for: All Gemini CLI developers — migrate to Antigravity CLI immediately
Workspace — final week of promotional access
The promotional open access for "Match Writing Style" and "Match Doc Format" features in Google Docs is entering its final week, set to expire on June 1, 2026.
Promotional access for Match Writing Style and Match Doc Format in Google Docs expires June 1, 2026. Paid plan required after that date.
Best for: Google Docs users — check your plan before June 1
Project Genie — early sandbox access
Developers subscribing to the highest AI Ultra tier ($200/mo) have begun receiving early sandbox access to construct interactive 3D environments from Street View data.
Project Genie sandbox access now rolling out to $200/mo AI Ultra subscribers. Interactive 3D environment construction from Google Street View data.
Best for: AI Ultra ($200/mo) subscribers — early access to 3D environment tools
Microsoft Copilot
Dateline: May 29, 2026 | Next update: June 5, 2026
The biggest story of the week is connectivity and workflow transparency: Microsoft introduced Federated MCP Connectors, expanded Excel guardrails, and added assistant-style features in Teams.
Federated MCP Connectors — live external data
Copilot can now query external systems live at prompt time via Federated MCP Connectors. Partners include Canva, HubSpot, Linear, LSEG, Moody's, Notion, and Google Calendar/Contacts.
Live queries via MCP-based connectors at prompt time — instead of pre-indexed snapshots. Enabled by default in tenants. Partners: Canva, HubSpot, Linear, LSEG, Moody's, Notion, Google Calendar/Contacts. Admin staging available via Microsoft Entra ID groups.
MCP-based connectors | Live queries instead of pre-indexed snapshots | Enabled by default in tenants | Admin staging via Microsoft Entra ID groups
Best for: Enterprises needing real-time data integration across apps
Excel — Plan Mode + Python scripting
Plan Mode shows users what Copilot will do before edits apply, adding transparency for analysts. Python scripting in Copilot for Excel is also live, enabling advanced data manipulation.
Plan Mode: preview multi-step changes before they apply — guardrails for analysts before workbook edits. Python scripting: integrated runtime now live for advanced data manipulation in Excel.
Plan Mode: preview multi-step changes | Python scripting: integrated runtime | Guardrails for analysts before workbook edits
Best for: Analysts, finance teams, data scientists
Teams — Copilot Call Delegation
Copilot can now answer incoming Teams calls, gather context from callers, and offer to book follow-ups.
Copilot call delegation workflow: answers calls, captures context, schedules follow-ups. AI usage policy updates recommended for enterprise deployments.
Call delegation workflow | Context capture + scheduling | AI usage policy updates recommended
Best for: Enterprise teams, executives, customer support
Mobile app — chat-first redesign
Copilot mobile app redesigned with chat-first interface and "liquid glass" styling.
UI overhaul with chat-first workflows and liquid glass styling. Faster access to connectors and notebooks from mobile.
Best for: Mobile-first professionals, on-the-go users
Security — Edge DLP redirect
When Purview DLP blocks a sensitive query to an external AI site, Edge now offers a "send to M365 Copilot instead" redirect.
Purview DLP integration: blocked sensitive queries now redirect to M365 Copilot. Requires tuned DLP policies to take effect.
Purview DLP integration | Redirect to Copilot | Requires tuned DLP policies
Best for: Enterprises enforcing strict data governance
Admin checklist — May 2026
Microsoft recommends five admin actions this month:
- Stage federated connectors before broad rollout
- Review Purview DLP policies for Edge redirect
- Pilot Plan Mode with analysts before comms
- Update Copilot training for Excel + Teams changes
- Revise AI usage policy for Teams call delegation
Plans and Pricing
No pricing changes this week. Federated connectors included in all tiers. Plan Mode and Python scripting general availability in Excel. Call Delegation enterprise only.