Latest
GuardFall Exposed Shell-Guard Bypasses in Ten Coding Agents
Adversa AI's June disclosure names the gap between the command your agent's safety filter reads and the command bash actually runs.
Read the briefing →Daily · Free · Independent
The Signal
Daily intelligence on AI, technology, and the decisions that follow.
Written for people who make decisions — not people who make predictions.
Get The Signal →Free to start. No spam. Unsubscribe anytime.
EU AI Office Enforcement Begins August 2
On August 2, Commission enforcement powers over GPAI providers begin, and Article 50 transparency obligations start to apply.
What Tokens per Watt Actually Measures
The metric compares token throughput with power draw, and comparisons require a common workload, latency target, and system boundary.
Chinese Models Take More Than 30% of US Companies' OpenRouter Tokens
DeepSeek V4 Flash lists at $0.14 per million input tokens, versus $5 for GPT-5.5, while Chinese models' share of US-company traffic on OpenRouter has remained above 30% since February.
Five New York AI Bills Await Hochul's Action
The package includes a one-year pause on state permits for data centers above 20MW and a ban on surveillance pricing based on personal data.
Gravitee: 14.4% of Surveyed Organizations Fully Approved Their Agent Fleets
A 919-person survey found full fleet approval was rare, while 59% reported confirmed incidents and 29% reported suspected incidents.
Meta Opens a Metered Muse API
Muse Spark 1.1 lists at $1.25 per million input tokens and $4.25 per million output tokens through Meta's proprietary model API.
FTC Frames AI Output Steering as a Deception Risk
The July 1 policy statement targets undisclosed model steering and sets up a fight with state AI-bias rules.
Washington Put GPT-5.6 Through a Pre-Release Hold
OpenAI's July 9 GPT-5.6 release followed a roughly 12-day federal review. The formal model-scope rule is due August 1.
Agent Gateways, Defined
Enterprise teams are putting one control point between autonomous agents, tools, credentials, and model calls.
What an agentic threat actor is
Sysdig coined "agentic threat actor" this week for JADEPUFFER, and the definition is already outrunning any page that pins it down.
An AI agent ran a ransomware attack with no human at the keyboard
Sysdig's JADEPUFFER writeup documents an autonomous agent that broke in, harvested credentials, moved laterally, and extorted a database start to finish.