One-Line Summary

OpenAI pushes ChatGPT deeper into everyday workflows — Codex arrives on mobile for on-the-go code reviews and a new personal finance preview links bank accounts — while Apple–OpenAI tensions add platform uncertainty.

New Tools

Codex comes to the ChatGPT mobile app

OpenAI is bringing its coding agent Codex into the ChatGPT mobile app so you can monitor and approve what it’s doing from your phone. The preview is available on iOS and Android, and after you connect to a Mac where Codex is running, you can review outputs, approve changes, and start new tasks; Windows support is expected later, per Reuters. ¹

OpenAI says the mobile experience goes beyond kicking off a single task: from your phone you can work across threads, review outputs, approve commands, change models, or start something new. The feature is in preview and available across plans, and arrives as Anthropic ships a similar remote capability for Claude Code. ²

Why this matters: phone-based control turns Codex from a desk-bound tool into a real-time control pane for teams, shortening the loop between a failing job and a fix. OpenAI also points to enterprise traction: Sea reports 87% of Codex users are weekly active and 73% would recommend it, framing Codex as more than autocomplete and closer to an agent that understands large codebases. ³

OpenAI’s education materials describe when Codex is the right fit: when you want an agent to operate on real files or support repeatable technical workflows — a pattern the mobile app now makes easier to supervise on the go. ⁴

ChatGPT adds a personal finance preview for Pro users

OpenAI is rolling out a personal finance experience in ChatGPT for Pro users in the U.S. on May 15, 2026, letting you securely connect bank and brokerage accounts, see a dashboard of spending and portfolios, and ask questions grounded in your own financial data. OpenAI says it supports more than 12,000 financial institutions and emphasizes user control over training settings, account disconnects, and data deletion within 30 days; it also notes ChatGPT is not a replacement for professional financial advice. ⁵

TechCrunch reports you connect accounts through Plaid, with Intuit support planned, and you can access the feature from a Finances entry in the ChatGPT sidebar or by typing “@Finances, connect my accounts.” The company highlights that more than 200 million people come to ChatGPT monthly for finance questions and plans to learn from Pro-user feedback before expanding availability. ⁶

OpenAI details how to remove connections in Settings and manage “financial memories” — saved context like goals or obligations — which you can view or delete. The feature is available on the web and iOS in this preview. ⁵

Industry & Biz

Apple–OpenAI tensions raise legal questions

CNBC reports that tensions between Apple and OpenAI could escalate into a legal battle, underscoring the stakes of platform relationships as AI features move closer to default mobile experiences. ⁷

Separately, OpenAI outlines updates to help ChatGPT better recognize context in sensitive conversations, including short-lived “safety summaries” that inform cautious responses in rare high-risk scenarios like self-harm or harm-to-others. OpenAI reports a 50% improvement in safe responses for suicide/self-harm cases and 16% for harm-to-others in long single-conversation tests. ⁸

For builders, OpenAI Support also posts guidance on ChatGPT App review rejections, a reminder that distribution now depends as much on policy fit and review criteria as on product features. ⁹

Community Pulse

Hacker News (467↑) — Readers like the idea of mobile control but raise setup friction and security concerns about blindly following system-level suggestions. ¹⁰

"For prompting an LLM specifically, what sort of formatting do you need, as you can just throw a stream of consciousness to it and it'll parse it just fine? For looking up other stuff, sure, but it's not too hard to open up new tabs in the mobile browser. I find it way more useful to be able to make stuff on the go rather than having to sit down every time. Does the tunnel setup feel clunky to use? That's the main thing that stops me." — Hacker News ¹⁰

"Also to drive this point home I've had claude suggest no password sudo before, some people are going to blindly accept that on a computer that also has access to accounts. Making that computer very vulnerable." — Hacker News ¹⁰

Hacker News (25↑) — Commenters welcome the personal finance feature but question liability if errors cause harm and whether reliability claims raise user risk. ¹¹

"Its fine if they offer that. But states need to make them liable for damages if it causes harm. The First Amendment protects them from criminal liability, but not civil. TOS' should not be able to shield them from the civil liability either." — Hacker News ¹¹

"OpenAI has advertised ChatGPT as a good and reliable source of information on everything" — Hacker News ¹¹

What This Means for You

• For product and engineering teams, Codex on mobile turns your phone into a control center for agent-driven work. That can cut response times when a job needs approval or a fix, but it also demands stricter guardrails — keep Codex on non-admin accounts, enforce MFA on the host machine, and treat phone approvals like code reviews, not rubber stamps. ¹

• For individuals and operations teams, the personal finance preview shows how general-purpose assistants become context-aware advisors once you connect data. If you try it, start with low-risk accounts, review data-sharing and model-training settings, and remember OpenAI’s caveat: ChatGPT is not a substitute for professional advice. ⁵

• For compliance and leadership, platform risk is not abstract: CNBC flags Apple–OpenAI tensions, and OpenAI is simultaneously tightening safety behavior in sensitive conversations. Contracts, app-store policies, and safety posture will shape which AI experiences are allowed to ship on phones you already use. ⁷

• For everyone experimenting at work, preview features change quickly. Document what your team approves from mobile, what requires desktop review, and how to disable access fast if an agent or tunnel behaves unexpectedly. ²

Action Items

1. Test Codex mobile on a safe repo: Update the ChatGPT app, link to a Mac running Codex, and practice reviewing outputs and approving one low-stakes change from your phone.
2. Create a “least-privilege” Codex host: Set up a non-admin user on a dedicated machine for Codex, require MFA, and explicitly disallow risky shell shortcuts (e.g., no password sudo) before enabling mobile approvals.
3. **Try the ChatGPT Finance preview ** (Pro, U.S.): Connect one low-risk account, turn off training if desired, ask three concrete questions (spending change, subscriptions, savings plan), then disconnect to confirm 30‑day deletion behavior.
4. Brief your team on sensitive-context safety: Read OpenAI’s safety update and add a one-page SOP for when to escalate, refuse, or reroute conversations that touch self-harm or harm-to-others.