OpenAI president Greg Brockman confirmed in a July 29 interview with Joanna Stern on YouTube that the company is developing a "family of devices" for interacting with its AI models, but declined to specify a release date, saying only that "you can expect them soon" (The Verge).

The interview, titled "OpenAI President On Reinventing Computers," is available in full on YouTube. The segment addressing hardware specifically runs from approximately 08:09 to 09:04. Brockman framed the device effort not as a single product but as a broader category of interaction surfaces designed around how he believes people will engage with AI systems.

A central thesis Brockman advanced is that voice will overtake typing as the dominant input modality. He said people will shift to "talking to computers rather than typing for the vast majority of what we do" (The Verge). That claim carries hardware implications: a voice-first interaction model demands always-listening microphones, on-device or low-latency inference, and acoustic environments that the current smartphone form factor only partially addresses.

On privacy, Brockman said OpenAI is working on "verifiable, auditable guarantees" that only the AI can review users' private data (The Verge). The phrasing suggests cryptographic or attestable mechanisms rather than policy commitments alone, though Brockman offered no architectural detail. For a company building consumer hardware with an always-on AI, the privacy posture is not incidental to the product. It is the product. Hardware that continuously captures audio or environmental data will face scrutiny from regulators and users alike, and "verifiable, auditable" is the vocabulary of zero-knowledge proofs and confidential computing, not terms of service.

Brockman was also asked about Apple's lawsuit against OpenAI. He deflected directly: "We are focused on our own development and technology" and "We have no interest in other companies' trade secrets. We are plenty innovative. We are thinking about things from our own angle" (The Verge).

The broader context here is that OpenAI is signaling a move from software and API services into physical computing devices, a category with fundamentally different economics, supply chains, and regulatory exposure. Brockman's description of a "family" rather than a single device suggests the company is exploring multiple form factors or tiers, though nothing in the interview confirmed specific designs, partnerships, or manufacturing arrangements.

The voice-first thesis is worth examining on its own terms. Speech recognition accuracy has improved dramatically, and large language models can now handle conversational disambiguation that earlier voice assistants could not. But voice as a primary interface also introduces new failure modes: ambient noise, multi-speaker environments, privacy concerns in public spaces, and the social awkwardness of talking to a device in a shared office or transit setting. Typing persists partly because it is silent, private, and reviewable before commitment. A shift to voice for the "vast majority" of computing tasks would require solving for all of those constraints, not just recognition quality.

The privacy guarantee language is the most technically interesting claim in the interview. If OpenAI is genuinely pursuing cryptographic attestation for data access, that would place it in the same conversation as confidential computing frameworks and secure enclaves, not merely data-use policies. Whether the shipping implementation matches the ambition remains an open question, and Brockman provided no timeline or technical specification to evaluate.

Brockman's deflection on the Apple lawsuit is notable for what it does not do. He did not deny the factual basis of the suit, address specific allegations, or discuss legal strategy. The statements about OpenAI's own innovation and lack of interest in others' trade secrets are general positioning rather than a legal response. That is consistent with a company that prefers to keep its public narrative focused on product vision rather than litigation.

For technology professionals, the key signals from this interview are three: OpenAI is building multiple hardware devices with no confirmed timeline, the interaction model is being designed around voice-first input, and the privacy architecture is being described in terms that imply cryptographic guarantees rather than policy commitments alone. Each of these carries different implementation timelines and different risk profiles, and none has yet moved beyond the stage of public statements without shipping artifacts.

The full interview is available on YouTube.