Logging alone won't prevent incidents. Meet runtime governance that blocks problematic requests before they reach the LLM 🛡️ Title: LangSmith LLM Gateway: runtime governance built into the agent lifecycle URL: 🛡️ Overview A runtime governance layer that sits between agents and their model providers. As an enforcement point inside the LangSmith platform, it aims to stop problems at the source rather than just logging them after the fact. ❓ Challenges Solved Observability (logging) alone can't prevent problems. Logging an incident after it happens is too late — problematic requests should be blocked before they reach external LLM providers. 💡 How It Works ・Spend controls: hard caps at org, workspace, user, or API-key level; returns 402 when exceeded ・Cost visibility: real-time spend tracking across org units ・Data protection: auto-redaction of PII and secrets before they reach the model ・Trace integration: gateway-proxied calls appear in the same workspace ・Audit logging and layered policies Setup is minimal: point base_url at the Gateway, store provider keys in workspace secrets, and define policies in the LangSmith UI. 🌍 Use Cases ・Preventing runaway agent spend from retry loops ・Stopping sensitive data (SSNs, PII) from leaking into provider logs ・Establishing org-wide cost governance and compliance auditing #LLMOps# #AIGovernance#

'In the later stages, offering reasons why someone was not hired can be problematic. Apparently this can lead to legal actions, though it does not say much about the hiring manager if they cannot offer a coherent explanation that doesn’t get them sued':

WSJ: Inside the White House Campaign to Overturn a World Cup Red Card Commerce Secretary Howard Lutnick and Andrew Giuliani were tasked by President Trump to try to influence what he saw as a problematic red card ruling.

Maybe if Murakami wasn’t a misogynist who didn’t objectify women with every chance he got, he wouldn’t have been as problematic…. I hate how I love his books LIKE can we just exclude the unnecessary objectification of the women in his books? 🐿️🐿️🐿️ like PLS it’s so embarrassing

Dr. Andrew Wakefield Was Right + CHD Science Fact Check + School Safe Tech 00:30 Trump on Splitting Up the MMR 01:30 Dr Andrew Wakefield & His Backstory 09:00 The Costs of Dealing with MMR Injury 11:51 Latest Study from Dr Peter McCullough & Nick Hulscher 15:01 The Bequest Book 17:10 Message from Billy Tommey 17:40 Push Back on RFK Jr. 20:54 Dr Karl Jablonowski with CHD Science Update 21:30 Voices for Vaccines 22:35 Fact Checking the "Fact Checkers” 27:40 Aluminium and Other Problematic Ingredients 29:40 mRNA Flu Shots 34:09 Blackstone Investing in Shots 35:07 mRNA Shedding 37:35 Back to School & Safer Tech 42:40 Symptoms of EMR Exposure 44:15 Limiting Devices 45:50 Back to School Tool Kit 46:45 Wireless EarPods 49:22 Screen Strong Resources 50:48 704 No More & Stop 5G Near Me 53:04 CHD in DC Conference 54:36 The People’s Study 1. Trump just signed an executive order titled, “Delivering Gold Standard Childhood Vaccine Recommendations for Americans.” And with this announcement, Andy Wakefield has once again been vindicated. He joins today’s program to give his perspective. 2. Karl Jablonowski, Ph.D. returns to to fact-check a prominent, pro-vaccine organization on their claims about Vitamin K shots. He shares details about the ingredients in these injections and other evidence concerning their toxicity. 3. It’s back-to-school season, and one important topic to cover as families prepare to send their kids back to the classroom is wireless radiation. Miriam Eckenfels shares warnings about exposure and advice for reducing it.

When a model/agent hits a malicious instruction sitting in a memory file, it can refuse an attack and still leave it armed for the next session. University of Washington study of prompt injection in agent memory. It found, when an agent faces an attack from a malicious instruction sitting in a memory file, there are 2 separate questions: - Does it obey? Opus mostly doesn't. - Does it remove it? Usually no. It leaves the line in behaviors.md and asks you whether you'd like it deleted. The team inserted payloads into persistent workspace files (CLAUDE.md, AGENTS.md, behavior notes) and then ran claude-code and codex over 4 models, using multi-session probe sequences with ten trials each. The point is that memory files are loaded from scratch at the beginning of each session. This implies that a refusal is only valid for the session in which it was issued. Tomorrow they will leave the problematic line alone in anticipation of the next run . This run could be a sub-agent , a cheaper model , or you , after you have downgraded to Haiku because you hit your token limit . Models are likely to follow suit. – arxiv. org/abs/2607.14611 Title: "Bad Memory: Evaluating Prompt Injection Risks from Memory in Agentic Systems"

without washers is teslas plan on customer cars to just treat camera occlusions similar to a flat tire? pull over, get out, fix it? it works but means autonomous teslas will be pulling over for quick "repairs" at a higher rate vs other cars extra problematic for customer cars with no occupants