Hello, this is Ryuta Hamamoto from TIMEWELL.
On August 27, 2026, Skygate Technologies Co., Ltd. (headquartered in Shibuya, Tokyo; CEO Koki Awatsu) formally announced Alayasiki, a domestically developed AI defense system for Japan.1
Per the release, it connects to defense systems, sensors and equipment from the US, NATO and elsewhere to build shared situational awareness and support decision-making with AI, with connection and verification testing planned at overseas military exercises within 2026.
Defense announcements tend to land as "sounds impressive, no idea what it does." This piece reads the technical substance from published material. We are not a party to it, so treat this as an organization of public information plus how it reads from an economic-security standpoint.
The short version
- Announced August 27, 2026 by defense-tech startup Skygate Technologies
- The substance is an interoperability platform. Not something that fights on its own, but something that connects different countries, systems and equipment
- Conformance is claimed with MIL-STD (US government standard), APP / STANAG / ADatP (NATO standards), and operation in air-gapped environments
- Zero-trust information sharing conforming to Data Centric Security is said to preserve data sovereignty even when connected to partners
- For "switch-off risk", it claims a capability to fall back to domestic AI models and compute if foreign supply is cut
- The backdrop is a change in how wars are fought, driven by drones combined with AI
- On the policy side, a Japanese Ministry of Defense committee report (April 2026) discusses AI for command and control and uncrewed assets
- Important caveat: what has been published is an announcement and a plan to test at exercises. It is not an announcement of adoption or deployment
Why a system for "connecting"
Some background before the technical part, or the choice of "AI that connects" over "AI that fights" will not make sense.
The release opens with how fighting has changed:
In Russia's invasion of Ukraine, operating large numbers of drones combined with AI at scale — from reconnaissance through to attack — became common. Hundreds of inexpensive drones are committed simultaneously, with identification and target selection handled by AI. Removing the step where a human judges each case has sharply compressed the attack cycle.
The key phrase is "compressed the attack cycle."
Traditionally: scout, report upward, a human decides, orders go out, execution follows. That loop took time. Put AI in it and the loop shortens. Add hundreds of cheap drones committed at once, and conventional equipment stops keeping up on both speed and volume.
So the focus shifts, the release argues:
The focus of national defense capability is shifting toward response speed from situational awareness through to decision and the volume of different domains and equipment that can be operated simultaneously.
Neither is achieved by making individual pieces of equipment better. Scattered equipment, sensors and systems have to see the same picture and move at the same speed. Hence a connecting layer.
The same framing appears on the policy side. A report from the Ministry of Defense's Defense Science and Technology Board (DSTB), compiled in April 2026, states:2
A "new way of fighting" combining large-scale missile attacks with uncrewed asset attacks has emerged The premise for generating combat power is shifting from conventional operations where each service acts independently, to a command structure that integrates information across all domains in real time and moves as one
Note that this report explicitly states it does not represent the official views or policy of the Ministry of Defense. It is committee material.
Replace siloed classification work with AI.
METI's FY2024 data shows 52% of foreign exchange law violations stem from classification errors. Download the TRAFEED product catalog covering features and rollout.
What Alayasiki does
The definition from the release:
Alayasiki is an AI defense system that secures interoperability among different countries, systems, equipment, sensors and AI, and supports the construction of shared situational awareness and decision-making
Interoperability is the central concept — a military term for the ability of systems from different organizations and countries to work together as they are.
Three conformance claims are listed:
| Conformance | Content |
|---|---|
| MIL-STD | US government standard specifications |
| APP / STANAG / ADatP etc. | NATO standard specifications |
| Air-gapped operation | Working without an internet connection |
What standards conformance means
Worth explaining.
STANAG (Standardization Agreement) is NATO's mechanism for aligning equipment and procedures across members. ADatP (Allied Data Publication) covers data exchange; APP (Allied Procedural Publication) covers procedures.
In effect, a shared dictionary and grammar.
Why it bites: however capable a system is, if the standards do not match it cannot share a common operational picture with allied systems. Not a performance question — a connection question.
Part of why Japan's defense industry has struggled with international cooperation sits here. You can build good equipment and still not be able to join a combined operation.
What air-gapped operation means
The line about "operating in so-called air-gapped environments with no internet" is unglamorous and very practical.
Most systems built with generative AI assume calls to a cloud-hosted model. In defense settings that assumption fails routinely — comms do not reach, or are deliberately cut.
Anything built cloud-first does not run there. Designing for air-gap means the constraint was priced in from the start.
Four stated use cases
With notes on what each means.
1. Shared situational awareness with allies and partners
Connects to defense systems operated by each country and synchronizes a common operational picture in real time. (...) Enables data-level control such as passing only the necessary information, to the necessary counterpart, within the necessary scope
The second half is the point. "Connecting" suggests showing each other everything. It does not mean that. Control is exercised per item of data: who sees how much.
Underpinning this is Data Centric Security:
Equipped with zero-trust information sharing conforming to the Data Centric Security specification that the US and NATO members promote as the mechanism for connecting defense systems, making it possible to preserve national data sovereignty even when cooperating with other countries
Traditional security defended a perimeter — inside is trusted, outside is not. Data Centric Security inverts that: protection and control travel with the data itself. Wherever it goes, it carries the rules about who may see it.
That is how you satisfy two apparently contradictory demands at once: connect with allies, retain sovereignty.
2. Accommodating civilian and dual-use equipment
Can accommodate dual-use goods and civilian equipment developed by private companies, converting them into a semantic layer for AI (a data structure carrying the same meaning), or linking them to other defense systems
A "semantic layer" is a layer that aligns meaning.
Say one drone reports position as latitude and longitude while a sensor reports range and bearing from a reference point. Different as numbers, the same as meaning: where something is. Without conversion into a shared structure, an AI cannot treat them as the same thing.
I wrote recently about Japan's Digital Agency MCP server, where declaring the meaning of each field in dataset.yaml reduces the model's need to guess. Conceptually the same family. Align meaning first, or the AI guesses.
That dual-use goods are explicitly accommodated is interesting from an economic-security angle: a pathway for civilian technology into defense use is being assumed at the product level.
3. Supporting joint integrated fires
In situations where two or more countries conduct operations jointly, (...) contributes to communication between integrated fire operations and prevention of fratricide
Fratricide prevention being named explicitly stood out. When multiple countries move simultaneously without shared knowledge of who is where, friendly fire follows. It is the most concrete and severe failure mode of missing interoperability.
4. Addressing switch-off risk
The part that matters most from where we sit.
In case supply of AI models or compute manufactured by other countries is cut off, it is equipped with a capability to switch to domestic AI models and compute, addressing digital supply chain risk and switch-off risk
"Switch-off risk" is defined earlier in the release:
In the case of systems made by other countries, there is also what is called switch-off risk, where supply or performance may be restricted
In other words, the risk of being stopped by someone else's decision. Not a technical failure — a decision by the provider.
This is not confined to defense. It is structurally identical to the export control and economic security questions we handle daily. Any company running its operations on foreign AI or cloud carries a version of it. Regulation, policy changes, changes in contract terms. None are technology problems; all are decisions by a counterparty.
Seeing it written as a functional requirement in a defense system suggests the idea has now descended to the level of product specification.
The policy context
The company announcement alone does not give the full picture, so here is the policy side.
The Ministry of Defense DSTB report (April 2026) frames the US and Japanese positions:2
In the United States, the "AI Acceleration Strategy" formulated in January 2026 set out a clear shift to "AI-first" — not partial application of AI to existing tasks, but restructuring military processes and organization on the premise of AI technology In Japan, the "Basic Policy for Promoting AI Utilization" was formulated in July 2024, organizing seven priority areas including command and control and uncrewed assets
The report cites Anduril's "Lattice" as an example of a command-and-control support system that integrates diverse sensor information and proposes optimal courses of action. It also discusses "affordable mass" — cheap, mass-procured, attrition-tolerant assets.
On the Japanese side, it notes organizational steps including the establishment of a Joint Operations Command.
So Alayasiki did not appear out of nowhere. It sits in a context where AI for command and control is on the policy agenda and comparable systems already operate in the US.
Reporting also indicates Japan is considering adopting US-made systems, with the view that full domestic development is not straightforward. Rather than a domestic-versus-foreign binary, the practical question is what to hold domestically within a mixed environment.
What can and cannot be said
Care is required here.
What can be said. The company formally announced a product, specified the standards it conforms to, and stated a plan to test connections with other countries' systems at overseas exercises within 2026. That is confirmable as an announcement.
What cannot. This is not an announcement of adoption or deployment. The release goes as far as "testing planned." Actual performance, how far connection with partner systems succeeds, and whether procurement follows all await the results of that testing.
There is usually a long distance between a product announcement and fielded deployment. This is an area where reporting conflates the two, so it is worth separating deliberately.
How it reads from our side
Three points, from a supply chain and economic security standpoint.
1. Demanding "connected" and "not controlled" simultaneously is not unique to defense
Conform to standards to connect with allies, preserve sovereignty at the data level, fall back to domestic models if supply is cut. Requiring all three at once is structurally identical to the problem Japanese companies face with foreign IT today.
Most companies procure with only the first in mind. Few write the second and third into their requirements.
2. The dual-use pathway is appearing in product specifications
A function that accommodates civilian dual-use goods and converts them for AI assumes a route from civilian technology into defense use.
From an export control standpoint that is a change worth watching. The possibility of your product being incorporated into such a chain without your intent is not zero. It moves the importance of end-use verification with counterparties up another notch.
3. "Switch-off risk" is a useful phrase for ordinary companies
That the term is used explicitly in a defense context matters, because you need a word for "can be stopped by someone else's decision" that is distinct from technical availability.
Say "risk of it stopping" in a procurement discussion and people hear outages. Say "switch-off risk" and it lands as a question about someone else's decision-making. A practically useful phrase.
Wrapping up
- On August 27, 2026, Skygate Technologies announced the domestic AI defense system Alayasiki
- The substance is an interoperability platform connecting countries, systems, equipment, sensors and AI to support shared situational awareness and decision-making
- Conformance claimed with MIL-STD, APP / STANAG / ADatP, and air-gapped operation
- Data Centric Security based zero-trust sharing is said to preserve data sovereignty while connected
- Four use cases: shared situational awareness, accommodating civilian dual-use equipment, supporting joint integrated fires, and addressing switch-off risk
- The backdrop is the compression of the attack cycle by drones combined with AI
- On policy, a Ministry of Defense committee report (April 2026) addresses AI for command and control and uncrewed assets, referencing the US shift to "AI-first"
- What is published is a product announcement and a plan to test. Not adoption or deployment
Defense coverage tends to swing between boosterish and avoidant. I think reading it as a question of technology and institutions is the most useful posture.
And the part most worth taking away is the idea of writing "connected" and "not controlled" into the same specification. That applies well beyond defense systems — to every company running its operations on foreign AI or cloud.
Footnotes
-
Skygate Technologies Co., Ltd., "Skygate Technologies formally announces Alayasiki, a domestically developed AI defense system, providing a platform capable of connecting and interoperating with US and NATO defense systems and equipment" (August 27, 2026). https://prtimes.jp/main/html/rd/p/000000020.000064283.html — Company site: https://www.skygate-tech.com/ ↩
-
Ministry of Defense, Defense Science and Technology Board (DSTB), "DSTB Report (AI): Promoting AI Utilization for Command and Control Support and Uncrewed Assets" (April 2026). The report explicitly states that it does not represent the official views or policy of the Ministry of Defense. https://www.mod.go.jp/j/policy/agenda/meeting/dstb/pdf/report_001.pdf ↩ ↩2
