AI news on Thursday, June 25, 2026

SIGNIFICANCE

★ Top story · Models2d ago

OpenAI Leans Toward Waiting Until 2027 for IPO, NYT Says

The Trump administration has directed OpenAI to stagger the release of its upcoming model, GPT-5.6, citing concerns over potential security risks. In response, the company plans to provide the technology to a select group of partners rather than launching it publicly. This intervention follows a similar move by Anthropic, which recently pulled its most advanced models due to regulatory pressure. The shift signals a new era of heightened federal oversight, where developers must prioritize restricted deployments over immediate, broad access to high-capability artificial intelligence.

TThe Decoder BBloomberg Technology BBloomberg Technology Read original ↗

Thursday — June 25, 2026

OpenAI Leans Toward Waiting Until 2027 for IPO, NYT Says

How agents are transforming work

Amazon Pledges Additional $13 Billion in India AI Outlays

Anthropic says Alibaba must be punished for largest Claude cloning attack

AI Demand Begins to Justify Massive Cost of Data-Center Buildout

Insurers turn to generative AI for catastrophe modeling, but hallucinations and sales logic could get in the way

DeepSeek on hiring spree – seeks newcomers, not just AI geniuses

DeepSeek Plans to Double Staff Across All Departments

Privacy-Aware Infrastructure in the AI-Native Era: An Asset Classification Case Study

Our latest Google Finance upgrades, including a new app

EU Held Talks With White House on Anthropic After Mythos Cutoff

Pentagon Sees Bigger Role for AI in Setting Military Targets

Authors Guild test finds some AI detectors perfectly identify human writing while others fail on every single text

Google bakes computer control directly into Gemini 3.5 Flash, letting the model see and operate your screen

Apple to Skip High-End M6 Mac Chips in Favor of AI-Focused M7 Line

TL++: Accuracy and Privacy Preserving Traversal Learning for Distributed Intelligent Systems

From Meta Idea to Advanced Mathematical Discovery -- Human-AI Co-Discovery of Sign-Embedding Quantum Algorithms

BiPACE: Bisimulation-Guided Policy Optimization with Action Counterfactual Estimation for LLM Agents

FDN: Interpretable Spatiotemporal Forecasting with Future Decomposition Networks

Structuring Sparsity: Block-Sparse Featurizers Capture Visual Concept Manifolds

How Modular Is a Frontier Mixture-of-Experts? A Pre-registered Causal Test in Which Apparent Expert Modularity Mostly Dissolves

Concept Removal for Frontier Image Generative Models

PatchINR: Patch-Based Implicit Neural Representations for Efficient and Scalable Inference

LLM-ACES: Closed-Loop Discovery of Dynamical Systems with LLM-Guided Adaptive Search

Black-Box Assisted Regression: Phase Transitions and Minimax Optimality

EPTS: Elastic Post-Training Sparsity for Efficient Large Language Model Compression

ShutterMuse: Capture-Time Photography Guidance with MLLMs

Semantic Allocation in Ordered Bottlenecks: Predictive Residual Inference for Visual Representation Learning

OPERA: Aligning Open-Ended Reasoning via Objective Perplexity-based Reinforcement Learning

GUI agent: Guided Exploration of User-Sensitive Screens

Dustin: Draft-Augmented Sparse Verification for Efficient Long-Context Generation with Speculative Decoding

What Actually Works for Spacecraft Fault-Tolerant Control: An Honest Settled-Gate Benchmark of Learned and Classical Methods

Diagnosing and Mitigating Compounding Failures in Agentic Persuasion via Taxonomic Strategy Retrieval

Do Thinking Tokens Help with Safety?

Conformal Orbit-Valid Trust Horizons for Equivariant World Models

What Do Language Priors Contribute to Darcy-Flow Inversion? A Mechanistic Audit

The Hitchhiker's Guide to Agentic AI: From Foundations to Systems

What's in an Earth Embedding? An Explainability Analysis of Location Encoders

Curvature-Guided Mixing for MLLM Adaptation

Holographic Memory for Zero-Shot Compositional Reasoning in Knowledge Graphs: A Mechanistic Study of Where and Why It Fails

From Forecasting Leaderboards to Deployment Decisions: A Fail-Closed Certification Protocol

Forget to Improve: On-Device LLM-Agent Continual Learning via Budget-Curated Memory

Cross-Modality Structural Guidance in 3D Latent Diffusion for Robust FLAIR Super-Resolution

A Spectral Phase Diagram for Binary Few-Shot Classification: Intrinsic Dimensionality, Geometric Saturation, and Representational Diagnosis

Chorus II: Cross-Request Sparsity Reuse for Efficient Image-to-Video Generation

Baidu Releases Unlimited OCR, a 3B Model That Keeps the KV Cache Flat for Long-Document Parsing

Scalable Peptide Design via Memory-Efficient Equivariant Transformer

Brevity is the Soul of Inference Efficiency: Inducing Concision in VLMs via Data Curation

What Intermediate Layers Know: Detecting Jailbreaks from Entropy Dynamics

ASAP: Agent-System Co-Design for Wall-Clock-Centered Auto HPO Research for ML Experiments

Constraint Tax in Open-Weight LLMs: An Empirical Study of Tool Calling Suppression Under Structured Output Constraints

Leaking Circuit Secrets: Gradient Leakage Attacks on Graph Neural Networks

Closed-Loop Graph Algorithm Execution with Small Language Models: Step Accuracy and Rollout Reliability

Minimalist Preprocessing Approach for Image Synthesis Detection

Neural Network Quantization by Learning Low-Loss Subspaces

Memory Makes the Difference: Evaluating How Different Memory Roles Shape Conversational Agents

Evidence for feature-specific error correction in LLMs

TRUSTMEM: Learning Trustworthy Memory Consolidation for LLM Agents with Long-Term Memory

MRI2Rep: Autoregressive Structured Report Generation for 3D Liver MRI

Dense Supervision Is Not Enough: The Readout Blind Spot in Looped Language Models

Neural Scaling Universality: If Exponents Are Fixed, Time to Understand Coefficients

UniTeD: Unified Temporal Diffusion for Joint Perception and Planning in Autonomous Driving

Internal Data Repetition Destroys Language Models

Retrieval-Augmented Personalization with Foundation Models for Wearable Stress Detection

Do vision-language models search like humans? Reasoning tokens as a reaction-time analog in classic visual-search paradigms

Falcon: Functional Assembly and Language for Compositional Reasoning in X-ray

LLM-Based Scientific Peer Review: Methods, Benchmarks, and Reliability Challenges

Latent Block-Diffusion Temporal Point Processes: A Semi-Autoregressive Framework for Asynchronous Event Sequence Generation

Transferability for General Reasoning: An Automated Curriculum for Multi-Domain RLVR

TACO: Towards Task-Consistent Open-Vocabulary Adaptation in Video Recognition

Causal-rCM: A Unified Teacher-Forcing and Self-Forcing Open Recipe for Autoregressive Diffusion Distillation in Streaming Video Generation and Interactive World Models

EchoStyle: Unlocking High-Fidelity Video Stylization with Reverse Data Synthesis

Dual Distribution Estimation for Zero-shot Noisy Test-Time Adaptation with VLMs

GCT-MARL: Graph-Based Contrastive Transfer for Sample-Efficient Cooperative Multi-Agent Reinforcement Learning

Yuvion VL: A Multimodal Foundation Model for Adversarial Content and AI Safety

Learning with a Single Rollout via Monte Carlo Pass@k Critic

The Generalization Spectrum: A Chromatographic Approach to Evaluating Learning Algorithms

TopoCast: A Topological Fidelity Framework for Evaluating Transformer-Based Time Series Forecasting

BitNet Text Embeddings