Skip to content
Daily Brief News Daily Brief

5 AI stories from June 3-7, 2026: Cursor SDK custom tools, OpenAI ChatGPT superapp plans, GPT-Rosalind biodefense update, Anthropic Project Glasswing expansion to 150 organizations, and Anthropic S-1 filing

· by Pondero Newsdesk · 5 stories

AI news daily brief: 2026-06-07

Five stories from June 3-7, 2026: one developer SDK release with affiliate relevance, one product strategy report citing more than a dozen sources, one model update targeting biomedical and government use cases, one enterprise security expansion, and one regulatory milestone signaling a coming IPO.

Cursor SDK ships custom tools, auto-review, and custom stores for headless agent workflows

Cursor published a changelog on June 4, 2026 covering four new capabilities in its TypeScript and Python SDKs, per the Cursor changelog.

Custom tools let developers pass function definitions via local.customTools on Agent.create() or per send(). The SDK exposes those functions through a built-in MCP server called custom-user-tools, removing the need to stand up a separate MCP server. Custom stores let teams replace the default SQLite backend with JSONL files, in-memory stores for CI environments, or any custom LocalAgentStore implementation. Auto-review routes headless tool calls through the same classifier Cursor uses in interactive mode; a permissions.json file steers which calls run immediately, which run sandboxed, and which require human approval. Nested subagents let parent agents spawn child agents to any depth, each with its own model, prompt, and tool set. The release also added run correlation IDs and checkpoint reliability fixes.

SDK pricing for headless runs has not been disclosed. The custom-tools feature currently covers local SDK runs only; extension to cloud agents has not been confirmed.

Read more: Custom stores, custom tools, and auto-review for the Cursor SDK. Try Cursor at cursor.sh.

OpenAI plans its biggest ChatGPT redesign since launch, bringing Codex and AI agents to the main interface

The Financial Times reported on June 7, 2026, citing more than a dozen current and former OpenAI employees, that OpenAI is planning its largest ChatGPT redesign since the product's launch. OpenAI has not confirmed the report. Reuters, which distributed the story, said it could not independently verify the account.

Per the FT, the overhaul gives Codex (OpenAI's coding agent, which the company said had 5 million weekly users as of June 3) greater prominence in the ChatGPT interface and bundles AI agent capabilities alongside partner integrations from Canva, Booking.com, and others. The FT framed the move as a response to Anthropic eroding ChatGPT's enterprise market share. The Decoder reported in March 2026, citing independent sourcing, that OpenAI was planning to merge ChatGPT, Codex, and its Atlas browser into a single desktop application under the direction of CEO of Applications Fidji Simo, per The Decoder. The two reports describe the same strategic direction from different time windows. No formal product announcement or UI launch date has been set.

Read more: OpenAI plans ChatGPT 'superapp' overhaul ahead of listing (Financial Times, June 7, 2026; paywall).

OpenAI updates GPT-Rosalind with agentic lab workflows and a free biodefense API program

OpenAI published "Introducing new capabilities to GPT-Rosalind" on June 3, 2026, per the OpenAI announcement. The update introduced three areas of improvement to the GPT-Rosalind model series.

Scientific reasoning was strengthened with grounding in real experimental literature. A multi-step workflow execution mode, which OpenAI calls "agentic lab workflows," lets GPT-Rosalind plan, execute, and revise sequences of wet-lab or computational steps. Access was expanded through two programs: the Rosalind Biodefense Program, which provides free API access to vetted developers and US government partners building biosurveillance, epidemiological modeling, and pandemic-preparedness tools; and a companion program described in a separate OpenAI post on societal resilience covering public-health use cases. The model is available via OpenAI's API on request. Pricing was not disclosed, and the model is not open-access.

Read more: Introducing new capabilities to GPT-Rosalind.

Anthropic adds 150 organizations in 15-plus countries to Project Glasswing for Claude Mythos security scanning

Anthropic published "Expanding Project Glasswing" on June 2, 2026, per the Anthropic announcement. Project Glasswing gives vetted organizations access to Claude Mythos Preview for autonomous vulnerability scanning.

The expansion adds approximately 150 new organizations based in more than 15 countries. The new cohort covers power, water, healthcare, communications, and hardware vendors. Anthropic noted that most new participants maintain codebases serving many downstream organizations, including infrastructure vendors and open-source maintainers. The initial cohort of approximately 50 organizations found more than 10,000 high- or critical-severity security vulnerabilities in production codebases, per TechCrunch. Separately, on June 1, Anthropic confirmed it gave the EU cybersecurity agency ENISA access to Claude Mythos through the same program. Each new participant must meet Anthropic's security requirements before gaining access. A timeline for making Claude Mythos broadly available beyond Project Glasswing has not been disclosed.

Read more: Expanding Project Glasswing.

Anthropic files confidential draft S-1 with the SEC for a proposed IPO

Anthropic announced on June 1, 2026 that it had confidentially submitted a draft Form S-1 registration statement to the US Securities and Exchange Commission for a proposed IPO of its common stock, per the Anthropic announcement. The announcement was made under SEC Rule 135. No share count, price range, or IPO date was disclosed.

The filing gives Anthropic the option to proceed with a public offering after the SEC review period, per the Anthropic announcement. Anthropic stated the offering will depend on market conditions and other factors. Bloomberg reported the filing within hours of the Anthropic post. Prior Bloomberg and TechCrunch reporting from May 20, 2026 indicated Anthropic reached its first profitable quarter in Q2 2026, with revenue surpassing $1 billion per month, and that the company carried a $65 billion post-money valuation after its Series H financing round. The Financial Times reported on June 7 that OpenAI is also preparing its own S-1, which would put the two companies in direct competition in public markets as well as in product.

Read more: Anthropic confidentially submits draft S-1 to the SEC.

Sources