Short-form institutional analysis
HSI Searchlight
Field signals and dispatches across the four research programs. Written for institutional readers who need to see what is changing and what is not yet visible.
Latest Issue · October 02, 2026
Issue 027
Companies Blaming AI for Layoffs Will Have to Say So on the Record
California has signed SB 951 into law. Its AI layoff disclosures will put a public legal record beside the explanations companies give investors and staff.
Read Issue 027 →Pillar
Email subscribers
Previous issues
27 issues published · Updated October 2, 2026
026
Pillar I · Responsible AI
The economics behind AI’s proposed slowdown: who gains from a truce, who sells the race, and where the costs fall.
September 25, 2026
025
Pillar I · Responsible AI
Anthropic has named its first marked Claude models and opened a detection API to an approved list. It has published an accuracy figure for neither. The commitment can be checked against Anthropic's own documentation, but the accuracy cannot be checked by anyone outside that list. Institutions drafting provenance clauses before December are writing rules around a capability they cannot test.
September 18, 2026
024
Pillar IV · Governance & Diplomacy
Interaction risk emerges in the handoff, when one level’s probabilistic output becomes the next level’s settled fact. Four episodes from India’s welfare, policing, and judicial systems show the uncertainty disappearing as decisions move upstream.
September 11, 2026
023
Pillar I · Responsible AI
The governable object in frontier AI has moved from what models can do to whether the organizations testing them can stop what they do.
September 4, 2026
022
Pillar IV · Governance & Diplomacy
AI risk is no longer contained within Big Tech. As hyperscalers spend past their own cash flow and turn to debt, the exposure is spreading into credit markets and institutional portfolios.
August 28, 2026
021
Pillar IV · Governance & Diplomacy
France just turned “AI summaries hurt publishers” into a formal competition-law test case, in front of a regulator that has already fined Google twice for this exact pattern.
August 21, 2026
020
Pillar I · Responsible AI
Professional licensure is becoming autonomous clinical AI’s first operational control layer. Utah is testing prescription renewals, the boards’ federation is drafting model text, and three incompatible licensure architectures are already on the table.
August 14, 2026
019
Pillar I · Responsible AI
A 1939 conjecture fell in a social post; twelve days later, ten machine-checkable results arrived for roughly $2,000 in compute. Mathematics is building the verification systems other institutions will need next.
August 7, 2026
018
Pillar II · Human Performance
Human oversight has become the default answer to AI risk. The 2026 evidence says the capacity for scrutiny that answer assumes is eroding, and eroding fastest among the people who use AI most.
July 31, 2026
017
Pillar III · Planetary Futures
Three hyperscaler climate reports in ten days show emissions climbing while the accounting is quietly re-engineered to keep net-zero pledges technically alive. Europe puts a September 27 deadline on the gap, and New York pulls the first statewide brake.
July 24, 2026
016
Pillar IV · Governance & Diplomacy
Tennessee’s SB 1580 made it illegal to market an AI as a mental-health professional — using a licensure definition and a 1977 consumer statute the state already had. Six states, three enforcement models, and the template most likely to spread.
July 17, 2026
015
Pillar I · Responsible AI
Banks increasingly run on judgment, values, and model access supplied by AI vendors — under usage policies, alignment choices, and political clearances they had no part in writing. Hilary Allen warns what that arrangement hollows out.
July 10, 2026
014
Pillar III · Planetary Futures
FERC put six regional grid operators on a sixty-day clock to justify or rewrite the rules governing how data centers connect to the grid — and the cost and speed of AI power now depend on the grid territory a deployment sits in.
July 3, 2026
013
Pillar IV · Governance & Diplomacy
With NIST yet to publish enforceable standards for AI agents, vendors are writing the rules faster than regulators can, and only a fraction of deployed agents have cleared formal security review.
June 19, 2026
012
Pillar IV · Governance & Diplomacy
The Commission's draft guidelines turn high-risk classification into a registered file that market surveillance authorities can demand on short notice, with the Article 6(3) filter as the line that decides it.
June 5, 2026
011
Pillar IV · Governance & Diplomacy
Four jurisdictions, four philosophies, no shared mechanism — and the cost of the gap falls on cross-border operators first.
May 22, 2026
010
Pillar I · Responsible AI
AI liability claims are landing on companies that assumed their existing coverage applied. At renewal, insurers want proof the AI is governed.
May 8, 2026
Archive
009
Pillar IV · Governance & Diplomacy
72% of enterprises say they govern AI. 43% can name the owner. The S&P 500 discloses AI risk at 83% and seats AI expertise on 2.7% of boards.
April 24, 2026
008
Pillar I · Responsible AI
Anthropic moved its largest customers to per-token billing the same week Stanford documented the widest confidence gap yet between AI experts and the public.
April 17, 2026
007
Pillar I · Responsible AI
Meta launched a proprietary model this week. Every company that built on Llama needs a new plan.
April 10, 2026
006
Pillar I · Responsible AI
A federal court ruled AI safety restrictions are protected speech. With frontier models converging on performance, the procurement decision just shifted from capability to vendor risk posture.
April 3, 2026
005
Pillar I · Responsible AI
OpenAI's Sora collapse, MCP's infrastructure milestone, and NVIDIA's agent push reveal which AI bets survive contact with real economics.
March 27, 2026
004
Pillar I · Responsible AI
Adecco locks in unlimited agentic AI. Atlassian cuts 1,600 to fund its pivot. The pricing floor for enterprise AI is being set now.
March 13, 2026
003
Pillar I · Responsible AI
The agent platform layer is fragmenting along trust lines. The operator playbook for navigating what's next.
March 6, 2026
002
Pillar I · Responsible AI
Retool data says 35% of teams replaced SaaS with custom builds. The decision rubric operators need this week.
February 27, 2026
001
Pillar I · Responsible AI
Enterprise AI agents broke the SaaS pricing model. This briefing on what to build, buy, and avoid.
February 20, 2026
About
What HSI Searchlight does
Searchlight is HSI's editorial voice. Each issue draws a single tight thread, often grounded in a recent regulatory move, market signal, or institutional decision. The goal is not breadth. The goal is to surface one thing that matters and make sense of it.
Every issue is grounded in HSI's four research programs — Responsible AI, Human Performance, Planetary Futures, and Governance & Diplomacy — and contributes to the Institute's wider research record. The strongest signals migrate into Research Briefs and Horizon Scans for fuller treatment.
HSI Searchlight is editorially independent. Patrons and partners support HSI's mission but do not direct editorial judgment, review issues prior to publication, or receive editorial input.
Editorial Masthead
The team behind HSI Searchlight
Editor-in-Chief
David Lovejoy
Founding Executive Director
Managing Editor
Ashwin Telang
Contributors
Research & Editorial Associates
Subscribe
Receive HSI Searchlight by email.
Each new issue arrives in your inbox on the day it publishes.
Need help subscribing? Contact HSI
Delivered through beehiiv. No marketing. Unsubscribe at any time.