What really happens when you say no

Samsung Health recently started asking users to let their personal health information be used for AI training and modeling. The request was framed as optional, but the warning that appeared when someone declined or withdrew consent told a scarier story.

It looked like Samsung Cloud syncing would stop, and all health information stored in the account would be permanently deleted. Given that Samsung Health can hold years of sleep data, exercise logs, medication schedules, and even medical records, users had every reason to panic.

Now Samsung has issued a new in-app notice that clears things up. The company says information collected for AI development is handled separately from the data required to run Samsung Health’s core features. When you withdraw permission, only the data gathered specifically for AI training gets removed. Your regular health records stay put.

The confusing wording that sparked the backlash

The original warning didn’t explain this division clearly. It mentioned AI consent, cloud syncing, and permanent deletion in close proximity, which made it easy to assume your entire health history was on the line.

Samsung has acknowledged the problem and says it’s revising the notice to make the policy easier to understand. That’s a sensible move, especially when the request covers deeply personal information like sleep patterns, medication records, menstrual cycle data, and health measurements.

What SamMobile’s testing revealed

SamMobile decided to test what actually happens when you opt out. After withdrawing consent, Samsung Health kept syncing normally, and the Samsung Cloud sync setting remained active. So based on both Samsung’s clarification and real-world testing, refusing Samsung Health AI training won’t stop you from syncing or accessing the health information you need.

Why this matters for your privacy

This incident highlights a bigger issue: consent prompts for sensitive data need to be precise from the start. When deletion is mentioned, users shouldn’t have to guess what gets deleted.

Samsung deserves credit for responding quickly instead of leaving the warning unexplained. But ideally, users should never have needed an external publication to figure out what would happen to their records.

If you’re worried about your own data, here’s what to keep in mind:

• Opting out of AI training only removes data collected for that purpose
• Your regular health history stays in your Samsung account
• Cloud syncing continues even after you withdraw consent
• Samsung says it’s improving the wording of the notice

What to do if you already opted out

If you declined or withdrew consent and are now worried about your health history, you don’t need to do anything. Your records are safe. The only thing that gets removed is the data specifically gathered for AI development, which is managed separately.

For those who haven’t decided yet, the choice is yours. You can allow your data to be used for AI training, or you can decline without losing access to Samsung Health’s features. Just remember that Samsung Health data deletion only applies to the AI training dataset, not your personal health records.

This whole episode is a reminder that clear communication about data practices isn’t optional. It’s essential, especially when dealing with something as sensitive as health information. Samsung has now clarified its position, but the lesson applies to every company handling personal data.

The two-year wait is over — almost

Apple Intelligence has finally cleared China’s regulatory hurdle. The country’s Cyberspace Administration has registered the AI suite for use on iPhones, ending a nearly two-year standoff that kept the feature out of the world’s largest smartphone market.

The approval doesn’t mean you can flip the switch today. Apple and regulators still need to set a launch date. But the main obstacle is gone, and that’s a big deal for Apple’s fortunes in China.

It’s been a long road. Apple introduced Apple Intelligence in June 2024, and mainland China users have been watching from the sidelines ever since.

Alibaba and Baidu step up as AI partners

Apple isn’t going it alone. Two of China’s biggest AI players are now officially on board.

Alibaba confirmed that its Qwen model will power Apple Intelligence experiences across iOS, iPadOS, macOS, and visionOS for users in China. Qwen handles text understanding, image recognition, and content generation directly inside Apple’s operating systems.

Baidu also confirmed it’s working with Apple on features for Chinese iPhone users. What’s less clear is how the workload splits between the two companies. Each model’s exact role remains undefined.

There’s another wrinkle. The approval only covers iPhones. Chinese iPads, Macs, and Vision Pro headsets are still waiting — the two-year AI drought is only partially over.

Why two partners instead of one?

Apple rarely relies on a single vendor for critical technology, and China’s regulatory environment makes diversification smart. Alibaba brings strong consumer AI chops with Qwen, while Baidu has deep experience in search and language models.

Having two partners also hedges against any single point of failure — a practical concern given the geopolitical sensitivity around AI in China.

What this means for Chinese iPhone users

Apple’s support documentation says compatible devices already sold in China should activate the service once it officially launches. That’s good news for anyone who bought an iPhone in the last couple of years expecting AI features that never arrived.

The prolonged absence put Apple at a real disadvantage. Huawei, Xiaomi, and other domestic manufacturers have been shipping generative AI on their phones for a while now. Apple was losing ground on a feature that’s become table stakes in the premium segment.

Apple and Alibaba actually submitted features for review back in early 2025. The process stalled — a casualty of the broader geopolitical tensions between China and the United States. Now that the logjam has broken, expect a rapid rollout.

The bigger picture for Apple in China

China remains a critical market for Apple, but it’s been a rough stretch. Local competitors are aggressive, and the government has shown it can hold up foreign tech when it wants to.

This approval signals a thaw — at least when it comes to AI. Apple’s partnership with Alibaba and Baidu also keeps it aligned with Beijing’s preferences: working with domestic firms rather than imposing US-centric AI models.

For iPhone owners in China, the wait is nearly over. For Apple, it’s a chance to reclaim momentum in a market where it can’t afford to fall further behind.

Want to catch up on how Apple Intelligence works elsewhere? Check out our guide on Apple Intelligence features explained and the latest on iPhone AI rollout news.

OpenAI’s New Enterprise AI Agent Play: A Managed Service, Not a Self-Serve Product

On July 22, OpenAI announced Presence, a managed enterprise AI agent offering that breaks from the company’s usual playbook. You can’t just sign up online and start using it. Presence is delivered through a limited general availability programme, with deployments led by OpenAI’s own Forward Deployed Engineers and a handful of selected global systems integrators.

This is a notable shift for a company that’s built its business on API keys and seat licences. Presence is sold as a project, not a product. Each engagement starts with a single, focused job — resolving a billing dispute, handling an insurance claim, or clearing an employee IT service request.

The agent gets only the knowledge and system access that specific job requires. The customer writes the rules: what the agent can do, when it needs sign-off, and when a human takes over. After launch, Codex reads production sessions and escalations, then proposes changes that the customer’s team tests and approves before rollout.

The Labour Behind the AI Agent

OpenAI’s documentation is refreshingly candid about the work involved. Its help centre lays out a six-stage process: scoping business outcomes, security and privacy review, legal review, simulation and acceptance testing, staged rollout, and post-launch iteration. A Presence agent, it says, does not become production-ready simply by ingesting documents.

That honesty matters. Gartner has warned that more than 40% of agentic AI projects will be cancelled by the end of 2027. The failures, Gartner says, stem from governance, undefined business value, and weak operational discipline — not from model capability.

Almost everything Presence bundles is aimed squarely at that diagnosis. Simulations and graders test whether an agent reached the right outcome, followed policy, used its tools correctly, and escalated when it should — before anyone outside the company speaks to it. Guardrails intervene when an interaction moves past defined boundaries. Session records and action histories give reviewers something to audit. Escalation paths hand a person structured context rather than a cold transcript. New versions go out through controlled rollout with rollback.

Enterprises have spent two years learning that the hard part of a production agent sits in integration, permissions, and change management. A vendor that sends engineers to do that work is responding to what buyers have actually been failing at, rather than shipping another dashboard and calling the gap a customer problem.

Where the Constraint Sits: Forward Deployed Engineers

The trade-off shows up in the eligibility criteria. Access, OpenAI says, depends on workflow fit, implementation readiness, and available delivery capacity.

Delivery capacity is a consulting constraint. Software scales; engineers cleared into a bank’s core systems do not. The title Forward Deployed Engineer is borrowed from Palantir, where it describes staff embedded in customer operations for months at a time. The economics attached to it look nothing like the economics of metered inference.

By putting its own FDEs and named partners at the front of every deployment, OpenAI has stepped into the layer of the market occupied by the integrators it will also rely on to scale. That’s a workable arrangement while volumes are small — and a more complicated one later.

It also raises a question for anyone scoping a contract. When the model vendor is also the implementation partner, the lines of accountability for a policy misapplied in production need to be written down rather than assumed.

The Enterprise AI Agents on Display Are Still Early

OpenAI describes Presence as battle-tested. Its case for that language: the product was assembled from years of deploying agents with enterprise customers before it was packaged and named. The claim is about accumulated practice rather than time in market, and it’s a reasonable one to make.

The strongest single proof point is OpenAI’s own English-language phone support line, 1-888-GPT-0090. The company says the agent met or exceeded internal benchmarks for frontline human support within weeks, now resolves 75% of inbound issues without human assistance, and cut human handoffs by 15 percentage points in ten days through the Codex improvement loop.

Those are OpenAI’s figures, measured against OpenAI’s own grading criteria, on OpenAI’s own channel. The transparency is welcome, but the numbers are not independently verified.

The three named customers sit earlier in the cycle than the launch framing implies. BBVA is exploring voice support for everyday banking in Mexico. SoftBank is testing Japanese-language conversations. IAG is exploring support during high-demand events such as severe weather. Daniel Ordaz, head of AI transformation at BBVA Mexico, describes the bank as a design partner helping shape and refine voice experiences for financial customer service. Design partners are normal and useful at limited GA. None of the three, though, is presented as running Presence at scale — worth holding alongside the word proven.

What OpenAI Hasn’t Disclosed About Presence

Pricing is not published. Implementation scope and cost are set per customer and per deployment. That’s ordinary for enterprise services, but it leaves buyers without a public reference point for cost per resolved contact against an incumbent contact-centre vendor.

The model is not named. Presence uses OpenAI models, the documentation says, with configuration selected for the workflow and subject to change as that workflow evolves. That flexibility is defensible engineering — pinning a production agent to a frozen model version ages badly. Teams that have spent the past year building evaluation suites against specific versions will nonetheless want the contract to say what they’re being held to when the configuration moves.

Channel support during limited GA covers voice or chat, with contact-centre integration, routing, authentication, and handoff design confirmed deployment by deployment. Data handling follows the same pattern, with the signed architecture and contract treated as the governing record rather than any published policy.

Presence sits apart from ChatGPT Workspace Agents, which remain the self-serve path for teams building inside ChatGPT and Slack. Voice customers keep API access to OpenAI’s frontier models. The company now offers broadly the same capability three ways, separated less by what the technology can do than by who does the work.

That leaves buyers choosing on delivery capacity as much as on model capability. And on OpenAI’s own account, delivery capacity is the part being rationed.

For a deeper look at how companies are putting agentic AI to work, see our coverage of HP accelerating enterprise workflows with OpenAI Frontier and the broader enterprise AI agent landscape.