AIUpdateWatch Daily Intelligence
A source-backed New York morning report focused on meaningful changes, with stable reference data kept on evergreen intelligence pages.
Executive summary
OpenAI began rolling out connected health information in ChatGPT for US users, Google published its first ATLAS study of how people use AI, DeepSeek reached the retirement deadline for two legacy API aliases, and Alphabet’s AI-capital-spending concerns helped pull technology shares lower.
The central shift is from experimenting with AI to governing it in sensitive daily use. Health data, workplace adoption, API migration deadlines, consumer expectations and the cost of AI infrastructure now require as much attention as model capability.
OpenAI launched Health in ChatGPT for US users, allowing eligible users to connect health information and let ChatGPT use it only under permissions they control.
Source (opens in a new tab)Google published the first ATLAS report, a research program examining how people use AI at work and in daily life, with the initial findings emphasizing assistance and task support rather than complete automation.
Source (opens in a new tab)DeepSeek’s legacy API names deepseek-chat and deepseek-reasoner reach their scheduled retirement on July 24 at 15:59 UTC; production systems should move to deepseek-v4-flash or deepseek-v4-pro.
Source (opens in a new tab)US markets recorded their worst daily loss in a month as Alphabet and Tesla fell after earnings and higher AI-related spending intensified concerns about capital requirements and cash flow.
Source (opens in a new tab)The FTC’s proposed AI-accuracy policy remains open for public comment through July 31, keeping disclosure, objectivity and consumer expectations on the near-term US policy calendar.
Source (opens in a new tab)Models
| Provider | Model | Availability | Modality | Best fit | Source |
|---|
Benchmarks
No new independently comparable cross-provider benchmark package was verified for the July 24 edition. The daily report therefore records operational, adoption, policy and market changes while stable scorecards remain on the evergreen benchmark page.
The methodology now explicitly tracks efficiency, latency and cost per successful task, but no unsupported speed ranking was created.
Concise status
Current pricing · Current United States hardware guide · Current open-model database
Markets
Quote timestamp: 2026-07-23T16:00:00-04:00.
| Item | Value |
|---|---|
| GOOGL | -7.1% |
| AMD | -2.3% |
| MSFT | -2.2% |
| NVDA | -1.6% |
| AVGO | -1.0% |
| SMCI | 2.2% |
| Ticker | Company | Close | Change | Source |
|---|---|---|---|---|
| GOOGL | Alphabet | $317.69 | -7.13% | Quote source (opens in a new tab) |
| AMD | AMD | $539.69 | -2.29% | Quote source (opens in a new tab) |
| MSFT | Microsoft | $381.58 | -2.24% | Quote source (opens in a new tab) |
| NVDA | NVIDIA | $208.76 | -1.56% | Quote source (opens in a new tab) |
| AVGO | Broadcom | $392.71 | -1.03% | Quote source (opens in a new tab) |
| SMCI | Super Micro Computer | $31.22 | +2.16% | Quote source (opens in a new tab) |
Regular-session snapshot. Quote providers may use different consolidation conventions. Informational only; not investment advice.
Industry
OpenAI says US users can connect supported health information, review synced records and control whether ChatGPT may use that context in a conversation.
High impactOpen original source (opens in a new tab)The first ATLAS report examines AI use at work and in daily life, helping separate observed behavior from broad automation claims.
High impactOpen original source (opens in a new tab)The deepseek-chat and deepseek-reasoner aliases are scheduled to stop working at 15:59 UTC. Explicit V4 model names remain the supported production path.
High impactOpen original source (opens in a new tab)Alphabet fell sharply after earnings as investors focused on capital expenditure and cash-flow pressure. The Nasdaq fell 2.2% in the broad selloff.
High impactOpen original source (opens in a new tab)The proposed policy statement addresses whether undisclosed manipulation of AI outputs could conflict with consumer expectations and federal deception rules.
Medium impactOpen original source (opens in a new tab)Research, safety and policy
Permission prompts are important, but users also need clear data controls, clinical limitations and guidance on when to seek qualified medical care.
Open original source (opens in a new tab)Teams should pin explicit model names, run regression tests, monitor errors and preserve a rollback path before the DeepSeek deadline.
Open original source (opens in a new tab)The proposal applies existing unfair-or-deceptive-practices concepts to AI representations and invites public comment through July 31.
Open original source (opens in a new tab)Next seven days
Traceability