LastWeekIn.Tech

AI policy coordination and safety governance dominate today's cycle, alongside enterprise integration deals and Apple's push toward provenance-verified photography.

1. 1

   Number 1. Policy & Regulation buttondown.com

   Anthropic, OpenAI, and Google coordinate on AI safety standards amid shifting geopolitical priorities

   Anthropic, OpenAI, and Google are reportedly coordinating on shared AI safety standards even as geopolitical priorities shift around them. The move comes amid broader industry debate over pacing frontier AI development and managing existential risk narratives.

   Why it matters Voluntary cross-lab coordination on safety standards is rare among fierce commercial rivals, and signals that leading labs see reputational and regulatory risk in appearing to race unchecked. If these standards translate into concrete evaluation or deployment commitments, they could become de facto industry norms before formal regulation catches up — but they could also be used to argue against binding government rules.

   What to watch Watch for whether these labs publish joint technical benchmarks or simply issue another vague joint statement, and how regulators in the EU and US respond.

2. 2

   Number 2. Security & Safety buttondown.com

   New safety startup led by former Anthropic and METR leaders raises $40m to rein in rogue AI agents

   A new startup called AIUC, founded by an early Anthropic hire and a former METR COO, raised $40 million to build technical solutions for controlling rogue AI agents. The company aims to develop tooling that reins in autonomous agent behavior before it causes harm.

   Why it matters As AI agents get deployed with real permissions across enterprise systems, the gap between agent capability and reliable control mechanisms is becoming a commercial and safety liability. Funding from credible safety-research alumni suggests investors see agent governance as a distinct, monetizable category rather than a feature bolted onto existing AI products.

   What to watch Look for AIUC's first product releases or partnerships with major agent platforms, and whether other safety-focused spinouts follow this funding pattern.

3. 3

   Number 3. Policy & Regulation techtimes.com

   Council of Europe moves forward on landmark AI privacy and LLM guidelines covering 55 nations

   The Council of Europe published a formal draft agenda advancing landmark privacy guidelines for AI chatbots and large language models under Convention 108, a treaty predating the internet that binds 55 nations beyond just the EU. The guidelines will address chatbot conversation retention, AI agent data-access scoping, and training-data extraction risks.

   Why it matters Because Convention 108 covers dozens of non-EU countries, this could create a much broader regulatory floor for AI privacy than the EU AI Act alone, affecting companies operating well outside Brussels' direct jurisdiction. The specific focus on training-data extraction attacks acknowledges a known technical vulnerability in LLMs that has mostly been treated as a research curiosity rather than a compliance issue.

   What to watch Track the Bureau's Consultative Committee review and whether member states begin drafting national implementing rules referencing these LLM-specific privacy standards.

4. 4

   Number 4. Companies aitoolsrecap.com

   Nvidia-backed Salesforce reasoning stack integrates into Claude across paid plans for thousands of sellers

   Salesforce's Nvidia-backed enterprise reasoning stack entered open beta inside Claude, available on all paid plans, bringing 37 prebuilt sales skills covering accounts, opportunities, and pipeline data. GitLab, Siemens, and Legora are already running the integration across roughly 7,000 sellers.

   Why it matters This marks a deepening of the Anthropic-Salesforce-Nvidia alliance to embed vertical enterprise AI directly into a general-purpose chat assistant, a strategy that could accelerate Claude's penetration into CRM-heavy sales workflows. It also raises unresolved questions about resource allocation — specifically whether this enterprise feature draws from the same capacity pool as consumer Claude usage, which reportedly saw allowances cut recently.

   What to watch Watch for Anthropic or Salesforce to clarify compute/capacity allocation for this integration, and whether other CRM or enterprise software vendors strike similar embedded-agent deals with foundation model providers.

5. 5

   Number 5. Security & Safety Hacker News

   Apple Reference Image: A New Approach for Verified Photography

   Apple has introduced 'Reference Image,' a new cryptographic approach to verified photography designed to distinguish real photographs from AI-generated or heavily altered images. The system responds to the growing difficulty of visually detecting synthetic images now that photorealistic AI generation tools are widely accessible.

   Why it matters As generative AI erases visual tells like malformed hands or garbled text, provenance-based verification — proving an image's chain of custody rather than analyzing its pixels — becomes the more durable defense against AI-driven misinformation. Apple's scale means this could become a de facto standard that other platforms and camera makers are pressured to adopt or interoperate with.

   What to watch Watch for adoption by news organizations, social platforms, or competing device makers, and whether standards bodies push for cross-platform provenance verification.