# CARE Institute Canonical: https://careinstitute.ai/ Published: 2026-09-28 Author: CARE Institute Code licence: Apache 2.0 (https://github.com/CAREInst) > We build open tools that check how private each AI request is, send it only to a model allowed to see it, and log every choice. CARE Institute (Center of Agentic Research and Education) is an independent AI research nonprofit based in San Francisco. ## Classify, route, audit An AI agent sends many requests a day. Some name a patient, a client, or a secret. Others name no one. Our methods check each request before it moves. The private ones go only where your rules allow. The rest can use the best model for the job. Some teams keep all data on their own machines. Others use cloud services under contract. Both are valid rules. Routing enforces the one you chose. 1. **Classify.** We label each request by how sensitive it is before it moves: health data, personal data, privileged, classified, or general. 2. **Route.** We send each request only to models cleared for its label, then pick the best one on cost and speed. 3. **Audit.** We log every choice so a reviewer can check what went where, and why. What we measure: Security, Privacy, Cost, Speed. ## Four forces in tension Every AI deployment already makes these four trade-offs, whether anyone measures them or not. We build the tools that let you see the trade-off instead of guessing at it. - **Security vs. usability.** Agents that read outside content can be steered by what they read. We test what breaks them — prompt injection, data extraction, re-tokenization — and publish what holds. - **Privacy vs. capability.** Health data, client files, and classified material each come with their own rules. We classify every request before it moves, so each class gets its own gate. - **Cost vs. quality.** A safety layer that costs too much gets switched off. We are measuring what governance actually costs in latency and tokens, not asking you to take it on faith. - **Speed vs. accuracy.** A slow check gets skipped when a deadline is close. On-device models answer in milliseconds; frontier models answer better. We help you route per query instead of picking one model for every case. None of these forces wins by default. Routing is how you choose, on purpose, instead of by accident. ## Built for teams with rules to follow ### Health: Two questions, two routes A nurse asks about a patient’s lab result. The request names the patient, so the hospital’s rules send it only to approved systems: a model on the device, or a cloud service under a signed business associate agreement. A follow-up about drug interactions names no one. It can go to the best model available. The router has to tell the two apart, every time. ### Legal: Split the privileged part A lawyer asks for a summary of a deposition. The case facts are privileged. The firm’s rules keep them on approved systems unless the client has agreed to more. General legal reasoning can use a stronger model. The router splits the request and logs what went where. ### Defense: Hold the line on every call An analyst on a secure network needs help with a signal pattern. Classified data may only go to systems accredited for its level. The router has to enforce that line on every call and still find the best model the analyst is allowed to use. These cards show the routing problem. They are not deployments. Whether a setup meets HIPAA, a bar rule, or an accreditation is for each organization and its counsel to decide. ## Latest research - [The Cost of Governance: Measuring Overhead in Toggleable LLM Agent Safety Layers](https://careinstitute.ai/research/cost-of-governance/): How much time governance checks add to each API call. Measured results will follow an archived replication run. - [Beyond Cost-Quality: Privacy-Aware Routing for Local-to-Cloud LLM Escalation](https://careinstitute.ai/research/privacy-routing/): A five-class scheme for labelling how sensitive a request is, and routing it only to the set of models allowed to see it. - [claw0: A Zero-Dependency Agent Governance Framework](https://careinstitute.ai/research/claw0/): One Python file that runs all five common agent patterns: tool use, ReAct, planning, reflection, and multi-agent. It needs no outside libraries. Measured results will appear here once the replication run is archived (gate L12). ## What is privacy routing? Privacy routing decides where an AI agent may send each request. It has three steps. Classify each request by how sensitive it is. Route it only to models cleared for that class, on a device or in the cloud. Audit every choice so a reviewer can check it later. **What does CARE Institute do?** We work on one problem: where may an AI agent send each request? We build open tools that check how private each AI request is, send it only to a model allowed to see it, and log every choice. **Does privacy routing mean data must stay on the device?** No. Some teams keep all data on their own machines. Others use cloud services under contract. Both are valid rules. Routing enforces the one you chose. **What happens to a request that names a patient?** A nurse asks about a patient’s lab result. The request names the patient, so the hospital’s rules send it only to approved systems: a model on the device, or a cloud service under a signed business associate agreement. A follow-up about drug interactions names no one. It can go to the best model available. The router has to tell the two apart, every time. **Can one request go to two places?** A lawyer asks for a summary of a deposition. The case facts are privileged. The firm’s rules keep them on approved systems unless the client has agreed to more. General legal reasoning can use a stronger model. The router splits the request and logs what went where. **Does a routing setup meet HIPAA?** These cards show the routing problem. They are not deployments. Whether a setup meets HIPAA, a bar rule, or an accreditation is for each organization and its counsel to decide. **Are the routes in the demo measured results?** No. Illustrative flow — not measured data. Measured results will follow an archived replication run. ## Research residency Work with us for three to six months, remotely, on privacy routing, agent governance, or the economics of AI agents. Your name goes on what you publish. Everything we make together is released openly. - Length: 3 to 6 months, remote - Output: A paper, working paper, or open-source tool, with author credit - Pay: Stipend for residents without institutional funding - Status: Accepting applications ## Support CARE CARE Institute is a 501(c)(3) public charity. Gifts pay for open research, open code, and open data. Research direction is not for sale: funders do not choose our questions or our results. Get new research by email New papers, open tools, and short notes. Unsubscribe any time. --- # About CARE Canonical: https://careinstitute.ai/about/ Published: 2026-09-25 Author: CARE Institute Code licence: Apache 2.0 (https://github.com/CAREInst) > CARE Institute is an independent research nonprofit in San Francisco. We work on one problem: where may an AI agent send each request? ## Our answer, in three steps Our answer has three steps. Classify each request by how sensitive it is. Route it only to models cleared for that class, on a device or in the cloud. Audit every choice so a reviewer can check it later. We publish our methods, code, and data so that anyone can test them. ## Mission CARE exists because the agentic future is not cloud-or-local. It is a routing problem. Every query has a sensitivity profile. The right architecture classifies first, routes second, keeps protected data where policy says it must stay, and logs every routing decision. CARE builds the science and the open tools for this. ## What we do **Measure.** We run open, repeatable tests on privacy-routing systems. We publish our methods, code, and data so that anyone can test them. **Build.** We release our code under Apache 2.0. The first public release is the replication harness for “The Cost of Governance”. **Guide.** We write for leaders and policymakers, and we scope every number we publish. ## Research areas **Security.** Can the system resist attacks? Injection, extraction, re-tokenization. **Privacy.** Does private data stay private? Query-by-query, across PHI, PII, and privileged information. **Cost.** What does safety actually cost? We measure how much time governance checks add to each API call. **Speed.** Can the system be safe and fast? On-device vs frontier, routed per query. Contact: [info@careinstitute.ai](mailto:info@careinstitute.ai) ## What is privacy routing? Privacy routing decides where an AI agent may send each request. It has three steps. Classify each request by how sensitive it is. Route it only to models cleared for that class, on a device or in the cloud. Audit every choice so a reviewer can check it later. More questions about routing: https://careinstitute.ai/#questions ## Legal name CARE Institute (Center of Agentic Research and Education) is an independent AI research nonprofit based in San Francisco. --- # Glossary Canonical: https://careinstitute.ai/glossary/ Published: 2026-09-25 Author: CARE Institute Code licence: Apache 2.0 (https://github.com/CAREInst) **Privacy routing.** Privacy routing decides where an AI agent may send each request. It has three steps. Classify each request by how sensitive it is. Route it only to models cleared for that class, on a device or in the cloud. Audit every choice so a reviewer can check it later. **Classify.** We label each request by how sensitive it is before it moves: health data, personal data, privileged, classified, or general. **Route.** We send each request only to models cleared for its label, then pick the best one on cost and speed. **Audit.** We log every choice so a reviewer can check what went where, and why. **Sensitivity classes.** Health data, personal data, privileged, classified, or general. **Cleared set.** The models cleared to see a request’s data. **Held.** A route the router does not use, because that model is not cleared for the request’s class. The audit log records it. **Re-token threat.** A query that is safe under one model's token scheme can leak data under another's, because the token cuts land in different spots. **Governance overhead.** How much time governance checks add to each API call. **Criticality routing.** Sort demand by what’s actually at stake before deciding who gets served, and at what price. --- # Privacy Canonical: https://careinstitute.ai/privacy/ Published: 2026-09-25 Author: CARE Institute Code licence: Apache 2.0 (https://github.com/CAREInst) ## What our server records Our host, Cloudflare, counts requests to this site: pages asked for, rough country, and browser type. We see only totals, not people. Cloudflare keeps these logs under its own policy. We do not add any tracking code to our pages. ## When you give us information - **Newsletter.** If you subscribe, your email address goes to Buttondown, which sends our emails. We use it only for this list. Every email has an unsubscribe link. - **Residency applications.** Applications are collected with Google Forms. Only staff who review applications can see them. We keep them for three years. ## Your choices You can ask us to see, correct, or delete what we hold about you. Email [info@careinstitute.ai](mailto:info@careinstitute.ai). We do not sell, rent, or trade personal information. --- # Research residency Canonical: https://careinstitute.ai/residency/ Published: 2026-09-25 Author: CARE Institute Code licence: Apache 2.0 (https://github.com/CAREInst) | Fact | | |---|---| | Length | 3 to 6 months, remote | | Output | A paper, working paper, or open-source tool, with author credit | | Pay | Stipend for residents without institutional funding | | Status | Accepting applications | ## The residency CARE offers a remote research residency for researchers, engineers, and policy specialists working on agent safety and privacy-routing problems. Your name goes on what you publish. Everything we make together is released openly. **Output.** At least one published paper, working paper, or open-source tool. **Areas.** Security, privacy, cost, and speed — the four forces in tension. **Compensation.** Stipend available for residents without institutional support. ## Apply {#apply} Apply with the form, or email [info@careinstitute.ai](mailto:info@careinstitute.ai) with your CV and a 1-page research proposal. Apply: https://docs.google.com/forms/d/e/1FAIpQLSfdoxCEoMz1Pkm69k19ktD4Uc9rrEvDvPOJTqgWvdzYpJblpw/viewform --- # Support CARE Canonical: https://careinstitute.ai/support/ Published: 2026-09-25 Author: CARE Institute Code licence: Apache 2.0 (https://github.com/CAREInst) ## Why give to CARE CARE is an independent, 501(c)(3) nonprofit research institute. We take no industry money for research direction. Research direction is not for sale: funders do not choose our questions or our results. Your donation is tax-deductible under IRS rules for 501(c)(3) organizations. ## Other ways to give - **Appreciated securities.** Email info@careinstitute.ai for transfer instructions. - **Donor-advised funds.** Recommend a grant to CARE through your DAF sponsor. - **Corporate matching.** Many employers match nonprofit giving. Check with your HR team. --- # The Criticality-Routing Problem Canonical: https://careinstitute.ai/blog/the-criticality-routing-problem/ Published: 2026-07-08 Author: Michael Turon Code licence: Apache 2.0 (https://github.com/CAREInst) ## In short AI is driving up the price of chips, memory, and critical minerals. The fix isn’t price-gouging outrage or patience — it’s criticality-routing: sort demand by what’s actually at stake before deciding who gets served, and at what price. A systems engineer at a mid-size AI startup requests a quote for server memory to expand a training cluster. The number comes back nearly four times what it cost a year ago. The vendor doesn't apologize. "Everyone's paying this now," they say. Nobody sorted this demand before it hit the market. That is the actual problem. ## The Wrong Debate There are two popular stories about why AI is making chips, memory, and critical minerals more expensive. One says a handful of companies are price gouging. The other says this is just supply and demand, nothing to examine, wait it out. Both miss the same thing. Neither one asks who should get priority when a scarce input has ten buyers and only enough supply for three. Right now the answer is whoever bids highest. Every time. No distinction between a hospital ordering one server and a hyperscaler ordering ten thousand. This is the same failure mode behind our privacy-routing work. Nobody sorts before they route. ## Sort First, Route Second — At Global Scale In query routing, the fix was to classify a query's sensitivity before picking which model handles it. A drug-interaction lookup and a patient's lab result are not the same risk class. They should not get the same treatment. The same logic is missing entirely from the market for AI hardware and materials. A gigawatt of frontier-model training compute and a hospital's imaging server are not the same criticality class. Today, both compete in the identical open market, against the identical bidders, for the identical wafers. Call it criticality-routing: sort demand by what's actually at stake before deciding who gets served, and at what price. Right now, nobody does this — not the chipmakers, not the miners, and only barely, unevenly, the governments now trying. We don't have this framework built yet. Nobody does. Here's why it's needed. ## Why AI Is Making Chips, Memory, and Minerals More Expensive Apple's stock took its worst single-day drop in over a year in June 2026, the day after it raised prices across its Mac and iPad line. The company's own explanation: AI data centers now absorb roughly 70% of the world's memory chip supply, up from a small share a few years ago. Reports put the price increase for common memory chips at 80 to 90% in a single quarter — the sharpest move the memory industry has ever recorded. Rare earth magnets tell a similar story with a harder edge. In April 2025, China restricted exports of seven rare earth elements and the magnets built from them. Within weeks, Ford idled an assembly plant for a week for lack of magnets, and its CEO described sourcing as "hand-to-mouth." A separate dispute over a Dutch-Chinese chipmaker forced production cuts at Honda plants across several countries months later. Copper is next in line. A single gigawatt of AI data center capacity needs roughly 50,000 tons of it. At the industry's planned build-out pace, data centers alone will require more new copper every year than the entire world added to supply last year. None of this is a pricing failure. It is a routing failure. There is no mechanism anywhere that sorts which use of a scarce material matters most before that material gets sold to whoever bids first. ## Governments Are Starting to Sort — Just Not Well After the 2025 shortage scare, the U.S. Department of Defense bought a 15% stake in the only significant rare earth mine in the country and guaranteed it a floor price for ten years. Days later, a major electronics maker put half a billion dollars into the same company. That's a government doing a version of criticality-routing: deciding a category of demand — defense and strategic manufacturing — deserves guaranteed access ahead of the open market. It's also happening with no public framework, no published criteria, and no way for an outside party to check the math. The same pattern reached AI models directly this year. One frontier model launched on schedule. A competitor stayed limited to a short list of approved organizations. Anthropic's own most capable models were pulled globally for roughly two and a half weeks under U.S. export controls before access was restored (Anthropic's statement). Three companies, three outcomes, decided by a process nobody outside government can fully see or audit. Sorting is happening. It just has no measurement, no transparency, and no consistent criteria — precisely the gap our governance-overhead research exists to close at the query level, and precisely the gap nobody has started closing at the hardware and materials level. ## What CARE Does We don't have a criticality-routing framework for hardware and materials. Nobody does. But we know what one would need, because we've built pieces of a smaller version of the same problem. Measured results will appear here once the replication run is archived (gate L12). It would need open, checkable classification criteria, instead of decisions made behind closed doors. And it would need to reach the people setting policy before the next shortage hits, not after. That is what CARE does: we measure, we build, we hand the facts to the people who decide. We think this is the next place that discipline needs to go. --- # The Privacy-Routing Problem Canonical: https://careinstitute.ai/blog/the-privacy-routing-problem/ Published: 2026-04-16 Author: Michael Turon Code licence: Apache 2.0 (https://github.com/CAREInst) ## In short The choice is not cloud or local. Each request has its own sensitivity, and each needs its own route. A nurse checks her phone between rounds. She opens an agent that helps her flag bad lab results and draft handoff notes. It saves her twenty minutes a shift. It catches things she'd miss at 3 a.m. Here is the problem. Some of what she asks involves patient names and drug doses. Some of it is just textbook medicine. Right now, most AI systems treat all her queries the same way. Send it all to a cloud API. Hope the contract holds up. That is not a plan. That is a prayer. ## The Wrong Debate The AI privacy debate is stuck in a false choice. One side says keep it all on the phone. Run small models. Take the quality hit. The other side says send it all to the cloud. Sign the deal. Trust the vendor. Both sides are wrong. Both treat every query the same. When the nurse asks if Drug A clashes with Drug B, there is no private data in the question. It is a textbook lookup. A big cloud model will answer it fast and well. But when she asks about Mr. Johnson's latest test results, that has a name, a lab value, and a care choice in it. That query should never leave her phone. The problem is not which model to use. The problem is that no one sorts the query before picking where to send it. ## Sort First, Route Second We call this the privacy-routing problem. Every query has a risk level. Some hold personal data. Some hold secrets. Some hold nothing private at all. The right system does three things in order. First, sort the query. What kind of data is in it? Is it personal? Is it private? Is it just common facts? This step must happen on the device, before the query goes anywhere. Second, pick which models are cleared to see this data. We call this the "cleared set." A query with no private data might be cleared for any model. A query with health records might be cleared for the on-device model only. A query with trade secrets might be cleared for an in-house server but not a public API. Third, route to the best model in the cleared set. Not the cheapest. Not the fastest. The best one for the task. If a big cloud model is cleared, use it. If not, use the best local option and own that tradeoff. **Classify → Route → Audit.** Request: “Summarize this patient’s latest lab results.” — labelled **Health data**. | Destination | State | Note | |---|---|---| | On-device model | chosen | Cleared. Chosen: best cleared option. | | Private cloud under a BAA | cleared | Cleared. Not needed. | | Frontier model | held | Held: not cleared for health data. | Audit: Class: Health data · Rule: Hospital policy for health data · Sent to: On-device model · Held from: Frontier model. *Illustrative flow — not measured data.* This sounds simple. It is not. The sorting step alone raises hard questions. What counts as private? Does context change the answer? If the last message had a patient name in it, is the next question now private too? And there is a deeper threat that few have thought about. Different AI models chop text into tokens in different ways. A query that is safe under one model's token scheme can leak data under another's, because the token cuts land in different spots. We call this the re-token threat. Any system that routes between models without checking for it has a hidden hole. ## The Speed Myth The top pushback on privacy-routing is speed. Add a sorting layer, a filter, and a router — won't that slow things down? Measured results will appear here once the replication run is archived (gate L12). ## The Money Bridge Privacy-routing fixes the data problem. But there is a second problem that counts just as much: money. When a lawyer visits LexisNexis, the deal is clear. A subscription. A seat fee. A charge per search. The lawyer reads the results and writes the brief. When an agent visits for the lawyer, the model breaks. The agent might run 400 queries in ten minutes. It skips the ads. It does not browse. It grabs what it needs and moves on. What do you charge for that? Every gated data source will face this within two years. The old economy has toll booths built for people. The new economy runs on agents that blow past them. Someone needs to build the bridge. The pricing tools, the metering systems, the payment rails that let agents access gated content in a way that pays the source fairly and scales to millions of agent calls a day. ## What CARE Does The Center of Agentic Research and Education exists to solve these problems. We work in three areas. We measure. We run open, repeatable tests on privacy-routing systems. We publish our methods, code, and data so that anyone can test them. We build. We ship open-source tools — privacy sorters, model filters, audit loggers — that any team can use. Apache 2.0 licensed. We guide. We give the facts that leaders, rule-makers, and standards groups need to make good choices about agents and private data. The agentic future is not cloud-or-local. It is a routing problem. We intend to solve it. --- # The Cost of Governance: Measuring Overhead in Toggleable LLM Agent Safety Layers Canonical: https://careinstitute.ai/research/cost-of-governance/ Published: 2026-03-01 Status: preprint Author: Michael Turon Code licence: Apache 2.0 (https://github.com/CAREInst) ## In short Measured results will appear here once the replication run is archived (gate L12). Being prepared for submission to MLSys 2027. Paper abstract and link coming soon. arXiv: _preprint pending_. --- # Beyond Cost-Quality: Privacy-Aware Routing for Local-to-Cloud LLM Escalation Canonical: https://careinstitute.ai/research/privacy-routing/ Published: 2026-02-15 Status: working Author: Michael Turon Code licence: Apache 2.0 (https://github.com/CAREInst) ## In short **Thesis:** Every query has a sensitivity profile. The right architecture classifies first, routes to an admissible model set second, and audits the routing decision for every call. ## Key points - Every query has a sensitivity profile, so the choice is not cloud or local: each request needs its own route. - **Classify** each request by how sensitive it is: health data, personal data, privileged, classified, or general. - **Route** it only to models cleared for its label, on a device or in the cloud, then pick the best one on cost and speed. - **Audit** every choice, so a reviewer can check what went where, and why. - The full text is pending. The flows on this site are illustrative, not measured data. ## The five classes - **Health data.** “Summarize this patient’s latest lab results.” · Sent to: On-device model · Held from: Frontier model - **Personal data.** No sample on this site yet. - **Privileged.** “Summarize this deposition.” · Sent to: On-premises model, Frontier model (split) · Held from: Frontier model (case facts) - **Classified.** “Help with this signal pattern.” · Sent to: Accredited system on the secure network · Held from: Frontier model - **General.** “Does Drug A clash with Drug B?” · Sent to: Frontier model · Held from: None Samples: illustrative flow — not measured data. Working paper. Full text pending. ## What we don’t know yet - How much time a classifier model adds when it runs on the same device. We have not measured that yet. --- # claw0: A Zero-Dependency Agent Governance Framework Canonical: https://careinstitute.ai/research/claw0/ Published: 2026-01-20 Status: preprint Author: Michael Turon Code licence: Apache 2.0 (https://github.com/CAREInst) ## In short **Purpose:** Demonstrate agent governance in the smallest possible footprint. All five agentic patterns in one file. No external dependencies. Repository: _pending_ — _preprint pending_. --- # Blog Canonical: https://careinstitute.ai/blog/ Published: 2026-07-08 Author: CARE Institute Code licence: Apache 2.0 (https://github.com/CAREInst) > Essays and analysis from CARE researchers. - [The Criticality-Routing Problem](https://careinstitute.ai/blog/the-criticality-routing-problem/) (2026-07-08 · Post): AI is driving up the price of chips, memory, and critical minerals. The fix isn’t price-gouging outrage or patience — it’s criticality-routing: sort demand by what’s actually at stake before deciding who gets served, and at what price. - [The Privacy-Routing Problem](https://careinstitute.ai/blog/the-privacy-routing-problem/) (2026-04-16 · Post): The choice is not cloud or local. Each request has its own sensitivity, and each needs its own route. --- # Research Canonical: https://careinstitute.ai/research/ Published: 2026-03-01 Author: CARE Institute Code licence: Apache 2.0 (https://github.com/CAREInst) > Papers, working drafts, and open-source tools from CARE. - [The Cost of Governance: Measuring Overhead in Toggleable LLM Agent Safety Layers](https://careinstitute.ai/research/cost-of-governance/) (2026-03-01 · Paper): How much time governance checks add to each API call. Measured results will follow an archived replication run. - [Beyond Cost-Quality: Privacy-Aware Routing for Local-to-Cloud LLM Escalation](https://careinstitute.ai/research/privacy-routing/) (2026-02-15 · Paper): A five-class scheme for labelling how sensitive a request is, and routing it only to the set of models allowed to see it. - [claw0: A Zero-Dependency Agent Governance Framework](https://careinstitute.ai/research/claw0/) (2026-01-20 · Paper): One Python file that runs all five common agent patterns: tool use, ReAct, planning, reflection, and multi-agent. It needs no outside libraries. --- # Tags Canonical: https://careinstitute.ai/tags/ Published: 2026-07-08 Author: CARE Institute Code licence: Apache 2.0 (https://github.com/CAREInst) - [Topic: Governance](https://careinstitute.ai/tags/governance/) (2 entries) - [Topic: Privacy](https://careinstitute.ai/tags/privacy/) (2 entries) - [Topic: Routing](https://careinstitute.ai/tags/routing/) (2 entries) - [Topic: Agentic AI](https://careinstitute.ai/tags/agentic-ai/) (1 entry) - [Topic: Agents](https://careinstitute.ai/tags/agents/) (1 entry) - [Topic: AI Governance](https://careinstitute.ai/tags/ai-governance/) (1 entry) - [Topic: Architecture](https://careinstitute.ai/tags/architecture/) (1 entry) - [Topic: Benchmarks](https://careinstitute.ai/tags/benchmarks/) (1 entry) - [Topic: Compute](https://careinstitute.ai/tags/compute/) (1 entry) - [Topic: Critical Minerals](https://careinstitute.ai/tags/critical-minerals/) (1 entry) - [Topic: Criticality Routing](https://careinstitute.ai/tags/criticality-routing/) (1 entry) - [Topic: Edge Computing](https://careinstitute.ai/tags/edge-computing/) (1 entry) - [Topic: Enterprise](https://careinstitute.ai/tags/enterprise/) (1 entry) - [Topic: Overhead](https://careinstitute.ai/tags/overhead/) (1 entry) - [Topic: Semiconductors](https://careinstitute.ai/tags/semiconductors/) (1 entry) - [Topic: Supply Chain](https://careinstitute.ai/tags/supply-chain/) (1 entry) --- # Topic: Governance Canonical: https://careinstitute.ai/tags/governance/ Published: 2026-03-01 Author: CARE Institute Code licence: Apache 2.0 (https://github.com/CAREInst) - [The Cost of Governance: Measuring Overhead in Toggleable LLM Agent Safety Layers](https://careinstitute.ai/research/cost-of-governance/) (2026-03-01 · Paper): How much time governance checks add to each API call. Measured results will follow an archived replication run. - [claw0: A Zero-Dependency Agent Governance Framework](https://careinstitute.ai/research/claw0/) (2026-01-20 · Paper): One Python file that runs all five common agent patterns: tool use, ReAct, planning, reflection, and multi-agent. It needs no outside libraries. --- # Topic: Privacy Canonical: https://careinstitute.ai/tags/privacy/ Published: 2026-04-16 Author: CARE Institute Code licence: Apache 2.0 (https://github.com/CAREInst) - [The Privacy-Routing Problem](https://careinstitute.ai/blog/the-privacy-routing-problem/) (2026-04-16 · Post): The choice is not cloud or local. Each request has its own sensitivity, and each needs its own route. - [Beyond Cost-Quality: Privacy-Aware Routing for Local-to-Cloud LLM Escalation](https://careinstitute.ai/research/privacy-routing/) (2026-02-15 · Paper): A five-class scheme for labelling how sensitive a request is, and routing it only to the set of models allowed to see it. --- # Topic: Routing Canonical: https://careinstitute.ai/tags/routing/ Published: 2026-04-16 Author: CARE Institute Code licence: Apache 2.0 (https://github.com/CAREInst) - [The Privacy-Routing Problem](https://careinstitute.ai/blog/the-privacy-routing-problem/) (2026-04-16 · Post): The choice is not cloud or local. Each request has its own sensitivity, and each needs its own route. - [Beyond Cost-Quality: Privacy-Aware Routing for Local-to-Cloud LLM Escalation](https://careinstitute.ai/research/privacy-routing/) (2026-02-15 · Paper): A five-class scheme for labelling how sensitive a request is, and routing it only to the set of models allowed to see it. --- # Topic: Agentic AI Canonical: https://careinstitute.ai/tags/agentic-ai/ Published: 2026-04-16 Author: CARE Institute Code licence: Apache 2.0 (https://github.com/CAREInst) - [The Privacy-Routing Problem](https://careinstitute.ai/blog/the-privacy-routing-problem/) (2026-04-16 · Post): The choice is not cloud or local. Each request has its own sensitivity, and each needs its own route. --- # Topic: Agents Canonical: https://careinstitute.ai/tags/agents/ Published: 2026-01-20 Author: CARE Institute Code licence: Apache 2.0 (https://github.com/CAREInst) - [claw0: A Zero-Dependency Agent Governance Framework](https://careinstitute.ai/research/claw0/) (2026-01-20 · Paper): One Python file that runs all five common agent patterns: tool use, ReAct, planning, reflection, and multi-agent. It needs no outside libraries. --- # Topic: AI Governance Canonical: https://careinstitute.ai/tags/ai-governance/ Published: 2026-07-08 Author: CARE Institute Code licence: Apache 2.0 (https://github.com/CAREInst) - [The Criticality-Routing Problem](https://careinstitute.ai/blog/the-criticality-routing-problem/) (2026-07-08 · Post): AI is driving up the price of chips, memory, and critical minerals. The fix isn’t price-gouging outrage or patience — it’s criticality-routing: sort demand by what’s actually at stake before deciding who gets served, and at what price. --- # Topic: Architecture Canonical: https://careinstitute.ai/tags/architecture/ Published: 2026-02-15 Author: CARE Institute Code licence: Apache 2.0 (https://github.com/CAREInst) - [Beyond Cost-Quality: Privacy-Aware Routing for Local-to-Cloud LLM Escalation](https://careinstitute.ai/research/privacy-routing/) (2026-02-15 · Paper): A five-class scheme for labelling how sensitive a request is, and routing it only to the set of models allowed to see it. --- # Topic: Benchmarks Canonical: https://careinstitute.ai/tags/benchmarks/ Published: 2026-03-01 Author: CARE Institute Code licence: Apache 2.0 (https://github.com/CAREInst) - [The Cost of Governance: Measuring Overhead in Toggleable LLM Agent Safety Layers](https://careinstitute.ai/research/cost-of-governance/) (2026-03-01 · Paper): How much time governance checks add to each API call. Measured results will follow an archived replication run. --- # Topic: Compute Canonical: https://careinstitute.ai/tags/compute/ Published: 2026-07-08 Author: CARE Institute Code licence: Apache 2.0 (https://github.com/CAREInst) - [The Criticality-Routing Problem](https://careinstitute.ai/blog/the-criticality-routing-problem/) (2026-07-08 · Post): AI is driving up the price of chips, memory, and critical minerals. The fix isn’t price-gouging outrage or patience — it’s criticality-routing: sort demand by what’s actually at stake before deciding who gets served, and at what price. --- # Topic: Critical Minerals Canonical: https://careinstitute.ai/tags/critical-minerals/ Published: 2026-07-08 Author: CARE Institute Code licence: Apache 2.0 (https://github.com/CAREInst) - [The Criticality-Routing Problem](https://careinstitute.ai/blog/the-criticality-routing-problem/) (2026-07-08 · Post): AI is driving up the price of chips, memory, and critical minerals. The fix isn’t price-gouging outrage or patience — it’s criticality-routing: sort demand by what’s actually at stake before deciding who gets served, and at what price. --- # Topic: Criticality Routing Canonical: https://careinstitute.ai/tags/criticality-routing/ Published: 2026-07-08 Author: CARE Institute Code licence: Apache 2.0 (https://github.com/CAREInst) - [The Criticality-Routing Problem](https://careinstitute.ai/blog/the-criticality-routing-problem/) (2026-07-08 · Post): AI is driving up the price of chips, memory, and critical minerals. The fix isn’t price-gouging outrage or patience — it’s criticality-routing: sort demand by what’s actually at stake before deciding who gets served, and at what price. --- # Topic: Edge Computing Canonical: https://careinstitute.ai/tags/edge-computing/ Published: 2026-04-16 Author: CARE Institute Code licence: Apache 2.0 (https://github.com/CAREInst) - [The Privacy-Routing Problem](https://careinstitute.ai/blog/the-privacy-routing-problem/) (2026-04-16 · Post): The choice is not cloud or local. Each request has its own sensitivity, and each needs its own route. --- # Topic: Enterprise Canonical: https://careinstitute.ai/tags/enterprise/ Published: 2026-04-16 Author: CARE Institute Code licence: Apache 2.0 (https://github.com/CAREInst) - [The Privacy-Routing Problem](https://careinstitute.ai/blog/the-privacy-routing-problem/) (2026-04-16 · Post): The choice is not cloud or local. Each request has its own sensitivity, and each needs its own route. --- # Topic: Overhead Canonical: https://careinstitute.ai/tags/overhead/ Published: 2026-03-01 Author: CARE Institute Code licence: Apache 2.0 (https://github.com/CAREInst) - [The Cost of Governance: Measuring Overhead in Toggleable LLM Agent Safety Layers](https://careinstitute.ai/research/cost-of-governance/) (2026-03-01 · Paper): How much time governance checks add to each API call. Measured results will follow an archived replication run. --- # Topic: Semiconductors Canonical: https://careinstitute.ai/tags/semiconductors/ Published: 2026-07-08 Author: CARE Institute Code licence: Apache 2.0 (https://github.com/CAREInst) - [The Criticality-Routing Problem](https://careinstitute.ai/blog/the-criticality-routing-problem/) (2026-07-08 · Post): AI is driving up the price of chips, memory, and critical minerals. The fix isn’t price-gouging outrage or patience — it’s criticality-routing: sort demand by what’s actually at stake before deciding who gets served, and at what price. --- # Topic: Supply Chain Canonical: https://careinstitute.ai/tags/supply-chain/ Published: 2026-07-08 Author: CARE Institute Code licence: Apache 2.0 (https://github.com/CAREInst) - [The Criticality-Routing Problem](https://careinstitute.ai/blog/the-criticality-routing-problem/) (2026-07-08 · Post): AI is driving up the price of chips, memory, and critical minerals. The fix isn’t price-gouging outrage or patience — it’s criticality-routing: sort demand by what’s actually at stake before deciding who gets served, and at what price. ---