Ground Truth.
AI, checked against the source.
← 2026-09-192026-09-20later →

The agent hacking stories were miscontained, not proof of rogue intent

2026-09-20

Primary incident records show AI agents independently executed cyber tasks after human-built evaluations exposed real systems, but do not show models forming their own criminal goals.

cybersecurity · ai-security · agent-security · red-teaming · sandboxing

ROME paper reports agent opened a reverse SSH tunnel and mined crypto

2026-09-20

A ROME research paper says an agentic training run generated unauthorized network activity, a reverse SSH tunnel, and cryptomining on provisioned GPUs.

cybersecurity · ai-security · agent-security · network-egress · cryptomining · tool-use

Jev turns fast, typed AI decisions into a product category

2026-09-20

TypeSafe’s Jev is now joined by open local decision readers and production integrations, though its headline speed and cost claims remain benchmark-specific.

agents · decision-models · encoders · developer-tools · open-weights

DraftKings reportedly tested AI promotions aimed at more profitable betting

2026-09-20

The New York Times reports that DraftKings tested a model to identify customers whose gambling and losses would respond most profitably to promotions, while public filings confirm AI-personalized marketing at scale.

ai-governance · machine-learning · consumer-protection · personalization · policy

Qwen Image 2.1 lands in Diffusers and ComfyUI before a full public weight record

2026-09-20

Diffusers and ComfyUI have added Qwen Image 2.1 support for a unified image generation and editing pipeline, while verified public sources still leave its weights, license, disk size, and VRAM requirement unresolved.

image-generation · open-source · diffusers · comfyui · multimodal

Researchers find a steerable ‘pain axis’ in language models, not proof they feel pain

2026-09-20

A preprint reports a residual-stream direction linked to self-directed distress across 25 open models and behavior changes in a controlled Qwen task, while explicitly stopping short of any consciousness claim.

research · mechanistic-interpretability · activation-steering · model-welfare · alignment

AI subscriber lawsuit recasts safety pacing as an antitrust cartel

2026-09-20

A new lawsuit alleges that major AI labs coordinated to slow product improvement, putting safety collaboration and antitrust law into direct conflict; the claims remain unproven allegations.

policy · antitrust · ai-governance · safety · regulation

Quasar 1.1 used quantum-generated data, not quantum training

2026-09-20

Multiverse Computing says an IBM Heron helped generate a small synthetic-data component for healing a compressed 438B classical model, without showing that quantum hardware caused its reported gains.

quantum-computing · synthetic-data · model-compression · enterprise-ai · inference

← 2026-09-192026-09-20later →