OpenAI Hardware Debut, Apple Document Demand, Bias Study

Three distinct OpenAI-adjacent stories broke in the last 48 hours: a first hardware product, an escalating legal dispute with Apple, and a new study questioning how frontier AI models handle political criticism.

Key points

• OpenAI has released its first branded hardware product, the Codex Micro keyboard, aimed at AI-assisted developer workflows.
• Apple sent preservation letters to roughly 40 former employees who now work at OpenAI, demanding documents and meetings with Apple lawyers.
• A Meta Oversight Board study found that OpenAI’s models, alongside those from other major labs, are less likely to generate criticism of governments that restrict free speech.
• A secondary market for used OpenAI-branded merchandise signals retail investor appetite for any form of proximity to the company ahead of a potential IPO.
• Check Point Software disclosed a partnership with OpenAI, using frontier models to simulate real-world cyberattacks against client infrastructure.

What is the Codex Micro keyboard?

OpenAI’s first hardware product is not the consumer AI companion device that many observers had anticipated from a company that acquired Jony Ive’s io design studio earlier this year. Instead, it is a compact keyboard built around Codex, OpenAI’s code-generation model, and positioned squarely at developers. The product reflects a deliberate choice to target professional workflows before tackling personal computing more broadly.

The move is notable for investors tracking OpenAI’s path toward hardware revenue diversification. A developer-first entry point keeps the initial market narrow, which limits near-term revenue upside, but it also sidesteps the high-stakes consumer launch risk that has historically tripped up other tech companies expanding into hardware. Whether this is a cautious opening or simply a staging product ahead of something larger remains unclear from current reporting.

How serious is the Apple legal pressure?

The Apple situation is a continuation of a story this site has covered, but the specific tactic disclosed today is new. Apple sent letters to approximately 40 former employees who now work at OpenAI, instructing them to preserve documents and make themselves available to Apple’s lawyers. Document preservation demands of this kind typically precede or accompany formal litigation. They signal that Apple is building an evidentiary record and wants to lock in materials before they are altered or lost.

The target is the individuals themselves, not OpenAI directly, at least at this stage. That distinction matters legally, but the practical pressure lands on OpenAI’s workforce. Employees receiving legal correspondence from a former employer face personal legal exposure, which can be distracting and expensive regardless of how the underlying dispute resolves. Apple has not publicly stated the precise claims it intends to pursue, and the source reporting does not elaborate on the alleged conduct at issue.

What did the political bias study find?

The Meta Oversight Board published a study examining how ten AI models from leading labs, including OpenAI and Anthropic, respond to requests for politically critical content across ten jurisdictions. The finding: models were measurably less likely to criticize governments known for restricting free speech, suggesting that training data or content policies are effectively echoing the speech rules of those countries.

The Oversight Board is funded by Meta but operates independently, and the study is its first focused on large language models. The methodology tested behavior across multiple geographies and models, including those from Meta itself, Google, and China’s DeepSeek. For OpenAI specifically, the finding adds to an ongoing debate about how commercially deployed AI systems balance global reach against political sensitivity. It is a reputational and regulatory consideration, particularly as OpenAI pursues enterprise and government contracts in markets where this kind of finding carries weight.

A cultural footnote on IPO anticipation

On the lighter end of the news cycle, Forbes reported that used OpenAI and Anthropic sweatshirts are selling on eBay for around $250. The phenomenon is trivial in financial terms but revealing as a sentiment indicator. When retail participants cannot access pre-IPO equity through normal channels, demand for symbolic proximity finds other outlets. It is a measure of how much anticipation has built around an OpenAI public offering, even as the company has provided no firm timeline for one.

Today’s edition is relatively quiet on the core IPO and fundraising front. No new valuation data, funding rounds, or governance changes were reported in the last 48 hours.