ChatGPT Saving Your Chats

Hello there! Ready to dive into another upgrading, Mind-boggling, and Value-filled Shift?

Today we have:

🛑 ChatGPT Logs Are No Longer Temporary, And OpenAI Is Pushing Back

🔥 How to Use ChatGPT’s Deep Research with Internal Connectors

🧠 Gemini 2.5 Pro Just Got Sharper, Again

🏆 Tools and Shifts You Cannot Miss

🛑 ChatGPT Logs Are No Longer Temporary, And OpenAI Is Pushing Back

A federal judge has ordered OpenAI to preserve all ChatGPT user logs indefinitely, including deleted chats and API outputs.  The order stems from the New York Times’ copyright lawsuit, but it could affect hundreds of millions of users worldwide.

The Shift:

1. The Court Order Covers Deleted, Temporary, and API Chats - The judge’s May 13 order requires OpenAI to retain all user outputs unless exempt by specific contracts, including deleted ChatGPT chats, temporary conversations, and API calls, even for standard users. Only ChatGPT Enterprise, Edu, and Zero Data Retention (ZDR) API customers are excluded.

2. OpenAI Says the Order Breaks Its Privacy Promises - OpenAI argues the ruling forces it to violate user privacy preferences and long-standing deletion policies. In filings and public statements, OpenAI claims the plaintiffs have no evidence of intentional data destruction, only speculative concerns.

3. The Stakes Are High for Business and Personal Users - OpenAI warns that preserving all chats could breach GDPR, violate contracts, and create security risks. Legal experts and users on LinkedIn and X have called the ruling “a breach of trust,” warning others to pause sensitive ChatGPT use.

4. NYT’s Case Is Based on Paywall Concerns, Not Proof - The lawsuit argues that ChatGPT could be used to bypass paywalls, and that users might delete chats to hide it. Judge Ona Wang said OpenAI might “never stop deleting” such evidence without intervention. OpenAI counters that there’s zero proof that users are abusing the system.

This legal precedent rewrites the rules of AI privacy. If upheld, it overrides OpenAI’s user-deletion controls and potentially exposes sensitive chat history to legal scrutiny. OpenAI is pushing back hard, but until the order is lifted, users, even those who deleted their chats, should assume everything is being saved.

🔥 How to Use ChatGPT’s Deep Research with Internal Connectors

Instant answers from Gmail, Drive, HubSpot, and more

ChatGPT’s new Deep Research + Connector feature is like having a personal analyst who searches your entire digital life—Gmail, Google Drive, HubSpot, Outlook, Dropbox, SharePoint, and more—in seconds. No more tab-hopping or digging through emails. Here’s exactly how to set it up and use it:

✅ Step-by-Step: Set Up and Use ChatGPT’s Deep Research

1. Make sure you’re a ChatGPT Plus or Enterprise user

• This feature is available in ChatGPT Pro (with GPT-4o) and Enterprise
  

• Go to chat.openai.com, click on your name (bottom left), then Settings → Beta features → Enable “Memory” and “Browse with web”

2. Connect your apps

• In the sidebar, click “Explore GPTs”, then search for “File Search” or “Connect apps”
  

• Click “Manage Connections” or go to chat.openai.com/a/connections directly
  

• Select and securely connect the tools you use:

  • Gmail

  • Outlook

  • Google Drive

  • Dropbox

  • HubSpot

  • Teams

  • SharePoint

  • GitHub

3. Start your Deep Research

Ask ChatGPT a specific question and it will instantly pull from your connected data.

Examples:

• “Summarize the last 3 emails from my investor.”

• “What were the Q4 campaign results in my HubSpot reports?”

• “What are the open action items in my Google Drive folders?”

ChatGPT will search across all connected sources in real-time and cite where each insight came from.

4. Refine or follow up naturally

Since it understands context, you can ask follow-ups like:

• “Summarize that into bullet points”

• “Generate a response email for that”

• “Create a slide based on those results”

Bonus Tips:

• You can filter results by source (e.g., “Only search Google Drive”)

• Works across internal docs, emails, slides, spreadsheets, CRM data, and more

It’s private and secure, only searching what you’ve explicitly connected

🧠 Gemini 2.5 Pro Just Got Sharper, Again

Google has released a new preview update of its Gemini 2.5 Pro model, now leading even more benchmarks and addressing previous performance gaps. This upgrade arrives just weeks after the I/O launch. It’s now available in the Gemini app, AI Studio, and Vertex AI with thinking budgets.

The Shift:

1. Top of the Leaderboards - Gemini 2.5 Pro now leads LMArena (Elo 1470) and WebDevArena (Elo 1443), continuing dominance in user-preference testing. It ranks #1 in image understanding (VibeEval), video understanding (VideoMMMU), and competitive STEM benchmarks like AIME 2025 and GPQA. 

2. Stronger in Coding, More Balanced in Style - The model continues outperforming peers on Aider Polyglot (82.2%) and shows improved performance in LiveCodeBench and SWE-bench. Google addressed criticism from May’s release by enhancing response formatting and fluency in non-coding tasks. 

3. Thinking Budgets & Dev Preview - Developers can now set "thinking budgets" in Gemini API to better manage latency and cost trade-offs. The preview is live in AI Studio and Vertex AI and will soon replace the current default Gemini 2.5 Pro version. 

4. Most Efficient on Price-Performance - Gemini 2.5 Pro is the most affordable across top-tier models, $1.25/M tokens for input and $10 for output, beating GPT-4o and Claude Opus on cost-efficiency. It also ranks top in multilingual reasoning (Global MMLU 89.2%) and code editing.

Google isn’t just building smarter models, it’s building smarter release cycles. With modular rollouts and public previews, it’s taking a faster, more transparent path to model iteration. Gemini 2.5 Pro now looks like the strongest general-purpose model on benchmarks and economics, and this drop just raised the bar again.

🔨AI Tools for the Shift

🤖 Rainex AI Agents – Automate digital business processes with powerful AI agents. Boost efficiency without writing code.

📝 ThinkFill.ai – Smart form filling that learns your writing style. Complete repetitive fields in seconds with AI memory.

💡 PromptVibe – Expert-level prompts, coding tips, and Cursor IDE rules to accelerate your productivity. High-quality resources that actually work.

🔍 Yahini – Replace AI fluff with actionable keyword strategies. Get funnel-mapped content briefs backed by advanced SERP and business analysis.

🖊️ fastwrite – Upload your research and get citable, AI-generated text suggestions. Seamlessly integrated with Microsoft Word.

🏃‍♂️Quick Shifts

🔍Anthropic has launched Claude Gov, a national security-focused AI model now used by top U.S. intelligence agencies. It loosens guardrails for classified data analysis and rivals OpenAI’s ChatGPT Gov offering.

❌X updated its developer terms to ban using posts for AI training, while still allowing select partners and its own AI, Grok, to train on platform data unless users opt out.

⚖️Meta’s Oversight Board says the platform allowed a deepfake scam ad featuring Ronaldo Nazário, viewed over 600K times. 

📃Anthropic CEO Dario Amodei opposes the proposed 10-year AI regulation moratorium, calling it too blunt. In a NYT op-ed, he urges a federal transparency standard instead of regulatory paralysis.

🤖Amazon has launched a new R&D group focused on agentic AI, housed within its hardware division Lab126. The team will develop frameworks to expand robotics capabilities, especially for warehouse automation.

That’s all for today’s edition see you tomorrow as we track down and get you all that matters in the daily AI Shift!

If you loved this edition let us know how much:

Forward it to your pal to give them a daily dose of the shift so they can 👇