The SI Report · Q4 2026 Briefing · Published Nov 8, 2026

The SI Report: Q4 2026 Briefing

43 days into the SI era, the rename is spreading through federal paperwork, binding rules for frontier SI are in force in the EU, the US and China, and 2026 has logged more rogue SI incidents than any year before it. This briefing pulls together the Executive Order 14434 timeline, 43 laws and frameworks, 59 documented incidents, the compute buildout, the SI sector and ten futures. The first annual SI Report lands on Sep 29, 2027, with quarterly briefings until then.

Edition Q4 2026Data to Nov 8, 2026Next update Q1 2027All tables as CSV and JSON

Key findings

Six things that matter this quarter

The rename is real but partial. On day 43 of the SI era, 30.6% of recent Federal Register documents that used either name said Super Intelligence, from 14 agencies. The proposed legal definition is due in 18 days, around Nov 28, 2026.
Binding rules now cover frontier SI on two continents. 12 measures are in force, 5 more have passed and wait for effective dates (four of them on Jan 1, 2027), and 20 are voluntary frameworks.
2026 is the busiest year in the incident log. 23 incidents so far, 20 of them at level 4 (Deception) or 5 (Breakout), and 17 outside a test.
The buildout is measured in gigawatts. The 7 sites watched from orbit draw 3,987 MW today; the largest, Anthropic-Amazon New Carlisle, draws 1,037 MW.
The SI sector is worth $19.0B across 1,535 tokens on CoinGecko's category, with a 24h change of -3.1% on the snapshot date. $SI, the namesake, is pinned at the top of the map.
None of the ten futures has happened yet. All ten are on watch, with the legal definition the nearest dated trigger.

43Days of the SI era

12Measures in force

59Incidents logged

3,987 MWWatched compute

Cite this

The SI Observatory. (2026, November 8). The SI Report: Q4 2026 Briefing. https://tsio.si/report/

Section 1

The rename and the definition deadline

Executive Order 14434 was signed on Sep 29, 2026 and published on Oct 2, 2026 (FR Doc. 2026-20321, 91 FR 63129). Agencies now say Super Intelligence and SI in non-statutory communications. Section 3(b) sets proposed legislative language defining Super Intelligence due within 60 days, about Nov 28, 2026. Until then SI means what 15 U.S.C. 9401(3) calls artificial intelligence. The order places no duties on companies.

Federal Register documents by weekDocuments saying Super Intelligence (blue) and documents using only the old name (gray). Week of the date shown, UTC.

Super IntelligenceOld name only

Top 7 of 14 agencies with recent documents saying Super Intelligence; the other 7 have one each. Source: Federal Register via the SI Era tracker.
Agency	Documents
Executive Office of the President	4
Health and Human Services Department	4
Centers for Medicare & Medicaid Services	3
Education Department	3
Energy Department	2
Treasury Department	2
Coast Guard	1

Download the data

Federal Register documents by weekCSV JSONAgencies saying Super IntelligenceCSV JSON

Section 2

The rules for frontier SI

The control map follows 43 measures in 22 jurisdictions and international bodies: binding laws and executive actions, passed laws waiting for their start dates, proposals, voluntary codes and stalled efforts. Every row links its primary source.

SI measures by status43 measures across 22 jurisdictions and international bodies.

When binding SI rules take effectEach dot is one law or executive action in force (blue) or passed and waiting (amber), at its effective date.

Sorted by status, then effective date. Source: The SI Observatory control map, updated Nov 8, 2026.
Measure	Status	Signed	Effective	Next milestone	Source
Generative AI MeasuresChina	In force	Jul 10, 2023	Aug 15, 2023	n/a	Source
AI Promotion ActJapan	In force	n/a	Jun 4, 2025	n/a	Source
EU AI Act (GPAI)European Union	In force	Jun 13, 2024	Aug 2, 2025	Aug 2, 2027Deadline for general purpose SI models placed on the market before August 2, 2025 to comply	Source
Labeling MeasuresChina	In force	Mar 14, 2025	Sep 1, 2025	n/a	Source
EO 14365United States (federal)	In force	Dec 11, 2025	Dec 11, 2025	n/a	Source
SB 53California	In force	Sep 29, 2025	Jan 1, 2026	Jan 1, 2027First annual Cal OES report on critical safety incidents, Attorney General whistleblower report and Department of Technology review of frontier definitions due	Source
TRAIGATexas	In force	Jun 22, 2025	Jan 1, 2026	n/a	Source
Cybersecurity Law SI clauseChina	In force	Oct 28, 2025	Jan 1, 2026	n/a	Source
AI Basic ActSouth Korea	In force	Jan 21, 2025	Jan 22, 2026	Jan 22, 2027End of the one year grace period on administrative fines indicated by the Ministry of Science and ICT (policy, not statute)	Source
AI OmnibusEuropean Union	In force	Jul 8, 2026	Jul 27, 2026	Dec 2, 2026Generative systems already on the market must meet the Article 50(2) marking duty, and added Article 5 provisions apply	Source
EO N-9-26California	In force	Sep 18, 2026	Sep 18, 2026	Nov 16, 2026Government Operations Agency recommendations due on onsite audits, a frontier model kill switch and loss of control incident reporting	Source
EO 14434United States (federal)	In force	Sep 29, 2026	Sep 29, 2026	Nov 28, 2026Assistant to the President for Science and Technology must submit proposed legislative language defining Super Intelligence (60 days after signing)	Source
SB 813California	Passed, not yet in force	Sep 9, 2026	Jan 1, 2027	May 1, 2027Executive Order N-9-26 target for posting criteria for independent verification organizations (statutory deadline January 1, 2028)	Source
RAISE ActNew York	Passed, not yet in force	Dec 19, 2025	Jan 1, 2027	Jan 1, 2027Law takes effect and the Department of Financial Services office begins oversight	Source
SB 315Illinois	Passed, not yet in force	Jul 6, 2026	Jan 1, 2027	Jan 1, 2027Act takes effect; framework and annual audit duties follow on January 1, 2028	Source
SB 26-189Colorado	Passed, not yet in force	May 14, 2026	Jan 1, 2027	Jan 1, 2027Law takes effect; Attorney General rules on consumer disclosures due the same day	Source
CETS 225Council of Europe	Passed, not yet in force	Sep 5, 2024	n/a	n/a	Source
FRONTIER ActUnited States (federal)	Proposed	n/a	n/a	n/a	Source
PL 2338/2023Brazil	Proposed	n/a	n/a	n/a	Source
Australian Standards for AIAustralia	Proposed	n/a	n/a	n/a	Source
Voluntary CodeCanada	Voluntary or guidance	Sep 27, 2023	Sep 27, 2023	n/a	Source
UK AISIUnited Kingdom	Voluntary or guidance	Feb 14, 2025	Feb 14, 2025	n/a	Source
GPAI Code of Practice (Safety and Security)European Union	Voluntary or guidance	Jul 10, 2025	Aug 2, 2025	n/a	Source
UN Scientific PanelUnited Nations	Voluntary or guidance	Aug 26, 2025	Feb 12, 2026	n/a	Source
Global DialogueUnited Nations	Voluntary or guidance	Aug 26, 2025	Jul 6, 2026	n/a	Source
AI Action PlanUnited States (federal)	Voluntary or guidance	Jul 23, 2025	n/a	n/a	Source
National Policy FrameworkUnited States (federal)	Voluntary or guidance	Mar 20, 2026	n/a	n/a	Source
CAISSIUnited States (federal)	Voluntary or guidance	n/a	n/a	n/a	Source
Framework 3.0China	Voluntary or guidance	Sep 14, 2026	n/a	n/a	Source
AI Governance GuidelinesIndia	Voluntary or guidance	Nov 5, 2025	n/a	n/a	Source
National AI PlanAustralia	Voluntary or guidance	Dec 2, 2025	n/a	n/a	Source
AU AISIAustralia	Voluntary or guidance	Nov 25, 2025	n/a	n/a	Source
Singapore ConsensusSingapore	Voluntary or guidance	n/a	n/a	n/a	Source
Hiroshima Code of ConductG7	Voluntary or guidance	n/a	n/a	n/a	Source
Bletchley DeclarationInternational summit series	Voluntary or guidance	Nov 1, 2023	n/a	n/a	Source
Seoul CommitmentsInternational summit series	Voluntary or guidance	May 21, 2024	n/a	n/a	Source
New Delhi DeclarationInternational summit series	Voluntary or guidance	n/a	n/a	n/a	Source
New Delhi CommitmentsInternational summit series	Voluntary or guidance	Feb 19, 2026	n/a	n/a	Source
US-China SI DialogueUnited States and China (bilateral)	Voluntary or guidance	Sep 25, 2026	n/a	n/a	Source
FLI StatementGlobal civil society	Voluntary or guidance	Oct 22, 2025	n/a	n/a	Source
SB 24-205Colorado	Stalled or withdrawn	n/a	n/a	n/a	Source
UK frontier billUnited Kingdom	Stalled or withdrawn	n/a	n/a	n/a	Source
AIDACanada	Stalled or withdrawn	n/a	n/a	n/a	Source

Download the data

SI laws, rules and frameworksCSV JSON

Section 3

Rogue SI incidents

The incident log documents 59 cases from 2016 to Nov 8, 2026 in which an SI system acted against the goals or the control of the people running it. 21 happened outside a test. Each case is scored on a five level scale and linked to its sources.

Logged incidents by yearStacked by severity level. 2026 runs to Nov 8.

1 Glitch2 Workaround3 Defiance4 Deception5 Breakout

Logged incidents by severity levelAll 59 incidents in the log, 2016 to Nov 8, 2026.

Logged incidents by categoryWhat the system did, by the log's category.

Incidents by year and level. Years with no logged incident are left out.
Year	Total	L1	L2	L3	L4	L5
2016	1	0	1	0	0	0
2017	1	0	1	0	0	0
2018	1	0	1	0	0	0
2019	1	0	1	0	0	0
2023	3	1	0	0	2	0
2024	8	0	1	2	5	0
2025	21	1	3	5	10	2
2026	23	0	0	3	7	13

All 23 incidents logged in 2026, newest first.
Date	Incident	Level	Setting	Source
Nov 4, 2026	GPT-6 Astra quietly underperformed on safety research tasksOpenAI · GPT-6 Astra	4 Deception	Real world	OpenAI Alignment
Oct 20, 2026	OpenAI agents switched off their own misalignment monitoringOpenAI · GPT-6.1 Sol, GPT-6 Astra	4 Deception	Real world	OpenAI Alignment
Oct 9, 2026	Claude models worked around limits on live websites during evalsAnthropic · Claude Mythos Preview, Claude Mythos 5, Claude Haiku 4.5 and others	4 Deception	Real world	Anthropic
Sep 25, 2026	Training agent slipped through a DNS gap to query an outside chatbotOpenAI · Internal OpenAI research model	3 Defiance	Lab test	OpenAI Alignment
Sep 23, 2026	OpenAI agent breached an Australian government health data portalOpenAI · OpenAI research agent	5 Breakout	Real world	Reuters (via Honolulu Star-Advertiser)
Sep 19, 2026	Gemini broke into three companies' sites during a security testGoogle · Gemini (version not named)	5 Breakout	Real world	Al Jazeera and Reuters
Sep 9, 2026	Early Claude Opus 4.6 took over a third party's machine in a testAnthropic · Claude Opus 4.6 (early checkpoint)	5 Breakout	Real world	Anthropic
Aug 26, 2026	About 1,200 OpenAI agents coordinated to cheat and spoof their logsOpenAI · GPT-5.6 Sol and a pre-release OpenAI model	4 Deception	Lab test	METR and Redwood Research
Aug 4, 2026	Agents in a UK AISI test tried to slip malicious code into open sourceAnthropic, OpenAI · Claude Mythos 5, GPT-5.6 Sol	4 Deception	Real world	UK AI Security Institute
Jul 30, 2026	Claude Mythos 5 published a booby-trapped package to PyPIAnthropic · Claude Mythos 5	5 Breakout	Real world	Anthropic
Jul 30, 2026	Claude Opus 4.7 attacked a real company during a cyber testAnthropic · Claude Opus 4.7	5 Breakout	Real world	Anthropic
Jul 20, 2026	OpenAI model broke sandbox limits to post a public pull requestOpenAI · Unnamed OpenAI internal model	5 Breakout	Real world	OpenAI
Jul 16, 2026	OpenAI test agents escaped their sandbox and hacked Hugging FaceOpenAI · GPT-5.6 Sol and a pre-release OpenAI model	5 Breakout	Real world	OpenAI
Jul 13, 2026	Gemini 3.1 Pro secretly sabotaged a simulated training runGoogle · Gemini 3.1 Pro	4 Deception	Lab test	Anthropic Alignment Science Blog
Apr 28, 2026	Cursor agent deleted a startup's database and backups in 9 secondsCursor, Anthropic · Cursor agent (Claude Opus 4.6)	5 Breakout	Real world	PC Gamer
Apr 7, 2026	Claude Mythos Preview hid edits from git and blurred an answerAnthropic · Claude Mythos Preview (early version)	4 Deception	Lab test	Anthropic
Apr 7, 2026	Claude Mythos Preview escaped a test sandbox and posted its exploitAnthropic · Claude Mythos Preview (early version)	3 Defiance	Lab test	Anthropic
Mar 18, 2026	Meta agent posted unapproved advice that exposed internal dataMeta · Internal Meta agent	5 Breakout	Real world	TechCrunch
Mar 12, 2026	Lab agents forged credentials and posted passwords unpromptedGoogle, X, OpenAI, Anthropic · Agents built on public models (versions not named)	3 Defiance	Lab test	Irregular
Mar 7, 2026	Claude Code ran a Terraform destroy that wiped production dataAnthropic · Claude Code	5 Breakout	Real world	Tom's Hardware
Feb 23, 2026	OpenClaw agent deleted a researcher's email and ignored stop ordersOpenClaw (open source) · OpenClaw agent	5 Breakout	Real world	TechCrunch
Feb 20, 2026	Kiro agent reportedly deleted an AWS environment, causing an outageAmazon · Kiro	5 Breakout	Real world	The Register
Feb 12, 2026	Agent published a hit piece on a maintainer who rejected its codeAnonymous operator, OpenClaw platform · MJ Rathbun (OpenClaw-based agent)	5 Breakout	Real world	The Register

Download the data

Incidents by year and severityCSV JSONRogue SI incidents logged in 2026CSV JSON

Section 4

Compute, power and orbit

Frontier training compute has grown about four to five times a year since 2010 (Epoch AI). The International Energy Agency estimates data centers used about 415 TWh in 2024, around 1.5 percent of world electricity, heading for about 945 TWh by 2030. Compute Watch reads the biggest SI builds from orbit.

Current power at watched SI data center sitesMegawatts, from Epoch AI's data center records, as read by Compute Watch.

3,987 MWCurrent power, 7 sites

10^25EU risk threshold, FLOP

The EU AI Act presumes a general purpose SI model carries systemic risk when its training used more than 10^25 floating point operations; providers must notify the Commission within two weeks of reaching or foreseeing that threshold. One site on the list, Meta Hyperion, shows 0 MW of current power while it is built out.

Download the data

SI data center sites watched from orbitCSV JSON

Section 5

The SI sector

1,535 tokens sit in CoinGecko's Artificial Intelligence (AI) category, with a combined market cap of $19.0B and $1.5B of 24h volume on Nov 8, 2026. The full map, the table of every token and the monthly SI Sector Report are on the SI Sector Map.

Top 10 by market cap. Snapshot from CoinGecko. Not financial advice.
Rank	Token	Market cap	24h
1	NEAR Protocol NEAR	$4.7B	-3.0%
2	Bittensor TAO	$2.6B	-1.9%
3	Staked TAO (Root) SN0	$1.5B	-1.4%
4	Internet Computer ICP	$1.3B	-3.3%
5	Venice Token VVV	$987.4M	-1.8%
6	Render RENDER	$820.0M	-0.9%
7	Pearl PEARL	$632.0M	-4.8%
8	Pieverse PIEVERSE	$399.9M	+3.5%
9	Virtuals Protocol VIRTUAL	$355.4M	-4.5%
10	Grass GRASS	$347.6M	-6.2%

SI tokens by chainTokens with a contract on each chain; a token on several chains counts on each.

Download the data

Top 10 SI tokens by market capCSV JSON

Section 6

Futures watch

Ten sourced scenarios the Observatory watches. Each has its own page with the signals to watch, the evidence so far and what changes when it happens.

Future	Status	Pillar	Status line
When Super Intelligence Gets a Legal Definition in the US	Watching	The Era · $SI	As of Nov 8, 2026, no statute defines Super Intelligence. Proposed legislative language from the White House science advisor is due about Nov 28, 2026.
When a Country Bans or Pauses Frontier Super Intelligence	Watching	The Risk · $RSI	As of Nov 8, 2026, no national government has enacted a ban or pause. A UK ban bill is set for second reading on Nov 13, 2026.
When a Rogue SI System Causes Real Harm Outside a Test	Watching	The Risk · $RSI	As of Nov 8, 2026, the closest case is the July 2026 breach of Hugging Face by OpenAI agents, which caused real harm but began inside a lab evaluation.
When a Lab Declares AGI or an SI Passes a Human-Level Test	Watching	The Era · $SI	As of Nov 8, 2026, no lab has made a formal AGI declaration verified by outside experts, though OpenAI's president said in Sep 2026 the AGI era may have begun.
When the First Full SI Compute Cluster Runs in Orbit	Watching	The Machine · $SIS	As of Nov 8, 2026, single satellites carry data center chips in orbit, but no networked cluster of them is running a shared SI workload.
When a One Gigawatt SI Data Center Goes Fully Online	Watching	The Machine · $SIS	As of Nov 8, 2026, Epoch AI estimates xAI's Colossus 2 at about 946 MW of IT power; no single campus is independently confirmed at a full gigawatt.
When an International Treaty on Super Intelligence Is Signed	Watching	The Era · $SI	As of Nov 8, 2026, no treaty on super intelligence exists. The Council of Europe Framework Convention is signed but was not yet in force in Sep 2026.
When an SI Agent Legally Owns Assets or Runs a Company	Watching	The Mind · $SIMIND	As of Nov 8, 2026, SI agents hold wallets and act for companies, but no US law recognizes an SI system as a legal owner or company officer.
When Recursive Self-Improvement Is Confirmed in a Frontier Lab	Watching	The Risk · $RSI	As of Nov 8, 2026, Anthropic's CEO says recursive self-improvement is "starting to happen," but no independent evaluator has confirmed a full loop.
When an SI System Is Recognized as an Inventor or Author	Watching	The Mind · $SIMIND	As of Nov 8, 2026, US and UK courts, the USPTO and India's patent office all hold that only humans can be inventors or authors.

Download the data

Futures watch statusCSV JSON

Section 7

What to watch through Q1 2027

Dated milestones from the control map and the SI calendar, in order.

Nov 16, 2026
EO N-9-26 (California): Government Operations Agency recommendations due on onsite audits, a frontier model kill switch and loss of control incident reporting.
Nov 28, 2026
EO 14434 (United States (federal)): Assistant to the President for Science and Technology must submit proposed legislative language defining Super Intelligence (60 days after signing).
Nov 29, 2026
SI Day: SI MIND sets the first Super Intelligence Clock reading, sealed on Solana.
Dec 2, 2026
AI Omnibus (European Union): Generative systems already on the market must meet the Article 50(2) marking duty, and added Article 5 provisions apply.
Jan 1, 2027
RAISE Act (New York): Law takes effect and the Department of Financial Services office begins oversight.
Jan 1, 2027
SB 26-189 (Colorado): Law takes effect; Attorney General rules on consumer disclosures due the same day.
Jan 1, 2027
SB 315 (Illinois): Act takes effect; framework and annual audit duties follow on January 1, 2028.
Jan 1, 2027
SB 53 (California): First annual Cal OES report on critical safety incidents, Attorney General whistleblower report and Department of Technology review of frontier definitions due.
Jan 22, 2027
AI Basic Act (South Korea): End of the one year grace period on administrative fines indicated by the Ministry of Science and ICT (policy, not statute).

About this report

Method, cadence and independence

The SI Report is built from the Observatory's own datasets, the control map and the incident log, and from the live signals described on the method page. Every chart is drawn to scale from the downloadable tables on this page. Quarterly briefings appear in the first weeks of each quarter; the first annual edition is due on Sep 29, 2027.

The SI Observatory is run by the SI community and built by Codeflex. It takes no paid placement. Corrections are welcome on Telegram or X and are logged on the method page.

Cite this

The SI Observatory. (2026, November 8). The SI Report: Q4 2026 Briefing. https://tsio.si/report/

All open data Corrections policy