01 / AI INFRASTRUCTURE

Alphabet AI Spending Commitments Lead Big Tech’s $3 Trillion Buildout

Alphabet has committed more than $1 trillion to AI infrastructure—part of a reported $3 trillion buildout across five Big Tech companies. The commitments cover buildings, equipment, power and leased facilities across different periods; they are not an annual budget or proof that all capacity is already in use.

02 / SOFTWARE PROTECTION

bla: add friction to AI reverse engineering of software artifacts

An open-source MCP and command-line tool designed to interfere with automated AI analysis of software artifacts.

Adds decoy logic and defenses aimed at automated analysis workflows.

Offers MCP and CLI integration; check compatibility with your build process.

Pricing and usage terms are unverified; test protection on your own artifacts.

03 / MODEL GOVERNANCE

OpenAI Model Spec Update Puts Human Intent at the Center of AI Judgment

OpenAI reorganized its Model Spec around five values: faithfulness to human intent, harm minimization, honesty and transparency, human agency, and care. The revision changes guidance and defaults as models take on more consequential actions.

04 / CODING AGENTS

Claude Code v2.1.314 Adds Firmer Boundaries for Automated Work

Claude Code v2.1.314 adds file-write restrictions across permission modes, group-based gateway routing and wider idle-conversation compaction. The release is about guardrails around automated coding work, not a claim of a new coding benchmark.

05 / SOFTWARE DEVELOPMENT

GitHub Octoverse 2026: Agents Start Over 40% of Pull Requests as Activity Surges

GitHub says agents supplied the first draft for more than 40% of September pull requests in its Octoverse 2026 report. That puts automated drafting inside the standard review workflow while leaving quality and acceptance as separate questions.

06 / OPEN-WEIGHT MODELS

Mistral Large 4 open weights put cyber defense under customer control

Mistral released downloadable Large 4 weights aimed at cyber-defense use cases. Open weights shift control of deployment and safeguards toward customers, along with responsibility for infrastructure, permissions and misuse controls.

07 / CLOUD COST CONTROLS

Google Cloud Spend Caps Reach General Availability, Trading Runaway Usage for Service Pauses

Google Cloud made spend caps generally available, with programmatic management, live estimated-cost reporting and monthly resets. Enforced caps can curb runaway usage but may also pause a workload when the limit is reached.

08 / AI SECURITY

Wiz Agentic Attacks Research Identifies Seven Fingerprints, With Limits to Attribution

Wiz described seven observable fingerprints of agentic attacks while warning that behavior alone may not identify the operator. The research separates containment signals from a definitive attribution claim.

09 / GOVERNMENT AI

ChatGPT desktop GovCloud deployment puts AWS controls at the center

AWS documented a ChatGPT desktop path for AWS GovCloud that routes inference through Amazon Bedrock and relies on workstation policy plus AWS identity controls. The guide describes a controlled evaluation route, not a transfer of approval responsibility.

10 / AI EVALUATION

Microsoft Copilot Synthetic Data Targets Enterprise Evaluation’s Privacy Gap

Microsoft described an enterprise evaluation approach that uses synthetic organizations instead of customer records to test Copilot. The goal is to reproduce workplace complexity without exposing confidential data.