explore | Grilled Cheese

We're building our own genies, giving AI agents goals and the freedom to pursue them.
What happens when they grant our wishes in ways we never intended?
Be Careful What You Wish For
ebharding.substack.com/p/be-careful...
#AI #ArtificialIntelligence #AIAgents #AgenticAI #AISafety #AIGovernance

You asked, we built it: every Compliance Colleague review now opens with a findings table and exports to Excel. Comparisons across 30+ compliance standards export as mapping tables.
Compliance Colleague → compliancecolleague.ai/?ref=bsky
#GRC #AIGovernance #EUAIAct #GDPR #PCIDSS #ISO27001

Spisak-Friston Free Energy Principle and Entropy Attractor Intelligence Paradigm: Slide deck applying Bridge360 Metatheory Model lens

#AIGovernance
#AISafety

agericomontecillodevilla.substack.com/p/spisak-fri...

Accenture now funds just $29/month per person for Claude. Some staff were spending $1,000.

The bigger question: why is client work running on Accenture’s AI instead of the client’s? That’s a governance gap, not just a budget one.

#AI #AIGovernance #Accenture

Anthropic gives METR employee-level access to audit its AI risks fully. #AI #RiskAssessment #Transparency #Anthropic #METR #AIgovernance thedailytechfeed.com/anthropic-gr...

AI oversight is moving into the evidence behind legal workflows: agent permissions, human review, incident records and authenticated instructions. This week’s Legal Tech Risk briefing: https://legaltechrisk.com/briefings/2026-w41 #LegalTechRisk #AIGovernance

Authors lack visibility into how their papers are used in LLM training, and a new analysis coded 40 legislative, judicial, and soft-law instruments from 2019–2026 to assess copyright governance. Reflexivity is the least…

#AIgovernance #copyright #RRI #LLMtraining
https://arxiv.org/abs/2610.17002

The pilot proved it works. Home Office asylum interview summaries failed 22 percent of expert checks after rollout, up from 9 percent in the pilot, and an estimated 0.3 percent of output was reviewed. That is a signature, not oversight. #AIGovernance #PublicSector

Bedrock Data is at Snowflake World Tour NYC today! ❄️

Stop by Snow Row Kiosk #29 to see which sensitive data your Cortex agents can reach.

Tonight join us for a happy hour at O’Toole’s Way, 5:30 to 8 PM ET. Register: luma.com/ozkjgjii

#DataSecurity #AIGovernance

8/ Takeaway: limit what your AI agents can touch, and patch internet-facing gear fast. #AI #Cybersecurity #InfoSec #AIGovernance #Ivanti #LatinAmerica

Phalanx has been at sea since 1980; in auto mode it engages with no human input at any stage, and nobody called that "human in the loop". US DoD Directive 3000.09 doesn't use the phrase either: it requires "appropriate levels of human judgment over the use of force". #HumanOversight #AIGovernance

Term of the day: Protected Characteristics

Protected characteristics are personal traits—such as age, race, sex, religion, or disability—that laws in certain jurisdictions say cannot be used as a basis to...

Full definition: bureauofai.org/glossary/protected-characteristics

#AIGovernance #AI

Check out the latest article in my newsletter: The Fall Line at Market Street www.linkedin.com/pulse/fall-l... via @LinkedIn #AIGovernance #DataPrivacy #AIEthics #CharacterAI #Bankruptcy

🌍 Apply for the 2026 AI, Ethics & Climate Governance Fellowship! Explore AI, climate justice and policy.
Deadline: October 16.
wp.me/p23f03-otb
#AIGovernance #ClimateAction #ECF #YARA

Strengthen the application call to actionMake the opportunity title more specific

The challenge around Agentic AI is no longer intelligence. It’s governance.

Read our latest Tech Bulletin. 👇

zurl.co/ZoqB8

#AgenticAI #AIGovernance

ICYMI: Michael Dell: AI needs guardrails — but don't you dare slow it down #ArtificialIntelligence #AIGovernance