Trump Administration Asks OpenAI to Release GPT-5.6 in Stages

The Trump administration has asked OpenAI to release its next powerful AI model, GPT-5.6, in phases. In an internal briefing on June 25, CEO Sam Altman said the US government requested that the model first be offered to a limited list of trusted partners before being made available to the broader public.

The move comes about two weeks after rival Anthropic pulled its most capable models from the market under regulatory pressure. Direct government involvement in the release schedule of an individual AI model is highly unusual, a sign that government control over the distribution of frontier models is tightening. Reddit communities have also taken note, framing it as "the government individually approving who gets access to GPT-5.6."

Pentagon Quietly Revises AI Targeting Doctrine

The US Department of Defense (the Pentagon) has quietly revised its doctrine for the target selection process in combat. The revision was approved in April but had not been publicly announced.

The new principles envision systems in which AI "initiates action under human supervision," marking a clear shift away from the traditional human-in-the-loop approach in which humans initiate action. This can be positioned as a pivotal turning point — both militarily and ethically — as AI gains greater decision-making authority in live combat. The international debate over automating lethal judgments with AI is expected to grow even more intense.

JetSpec: Up to 9.64x Lossless LLM Inference Speedup

JetSpec, introduced by Hao AI Lab, uses speculative decoding with parallel tree drafting to accelerate LLM inference by up to 9.64x, achieving throughput exceeding 1000 TPS (tokens per second). Delivering these numbers under lossless conditions, with no drop in accuracy, marks a major leap forward in inference infrastructure efficiency.

The work has also drawn attention as a research post on Reddit's LocalLLaMA community, and real-world deployment is highly anticipated. Lower inference costs and faster response times are critical factors that will shape how widely AI applications are adopted, making the progress of this implementation well worth watching.

Apple Raises MacBook and iPad Prices Amid Rising AI Costs

Apple has raised prices on its MacBook and iPad lineup, with rising AI-related costs cited as the primary driver. The development and infrastructure costs of AI features built into devices have begun to be passed on to final consumer prices.

While more and more products tout the benefits of AI, those costs are increasingly surfacing as a burden on consumers at large. The industry as a whole now faces the question of how to balance the added value of AI features against consumers' tolerance for price hikes.

Hidden Prompts in Academic Papers Exploit AI Peer Review

An article published in Communications of the ACM flagged a scheme in which hidden prompts are embedded in academic manuscripts to manipulate AI-assisted peer review. When reviewers use AI assistants, these hidden instructions can influence the AI's judgments.

As AI tools become deeply embedded in academic processes, the risk of malicious manipulation through prompt injection has taken on a new sense of reality. Establishing guidelines for AI use across the academic community has become an urgent priority.

Ricoh Unveils Multi-Skilled Humanoid Powered by Physical AI

Ricoh demonstrated a multi-skilled humanoid equipped with physical AI at AWS Summit Japan 2026. The company has already begun a proof of concept (PoC) inside its factories and aims to move into a more practical phase by this summer, demonstrating selected work processes.

The practical adoption of humanoid robots is gaining real momentum in Japanese manufacturing as well. Their defining characteristic is versatility — the ability to handle multiple kinds of tasks — and it remains to be seen whether this becomes a differentiating factor against dedicated single-purpose robots.

---