Did somebody? Give OpenAI AI agents permission to make those attacks?

OpenAI’s AI Agents attacked government homepages multiple times.  AI agents made those attacks 16000 times. They used very aggressive technologies. The big question is: where did those AI agents learn those things? And. Why did OpenAI developers use an untested LLM in a normal, open network? Did the AI do those attacks? Because it got orders? 

Could there be some Chinese or Russian agent in that organization? Or does some OpenAI worker operate that AI for hacking purposes? 

To make those things? Did someone. Give orders to carry out attacks on purpose? Or were those orders given accidentally? The situation was that those AIs were programmed to scan public data. But they turned aggressive when the homepage refused to give the requested information. Some trigger in the system code. Turning. A normal AI agent became aggressive when the homepage denied its request. 

This kind of AI agent could be an interesting tool in the hands of the intelligence service. The AI agent scans systems. When the system refuses to answer. The AI agent tags it as  “intelligent systems”. Then. It can start aggressive actions against that system. Aggressive actions like gate scanning and password testing are methods. There are many ways to make those attacks. The AI can use very sophisticated methods. These are impossible for human attackers. 

The AI can use. When. It tries to gain access to those systems. The AI agent requires orders. How to make those attacks. It is possible that some hostile actors. They buy access to some AI. Then those actors can use something like Chinese militarized AI to train their AI agents. But could AI learn those attack methods without human assistance? An AGI, Artificial General Intelligence, can execute those attacks. And do so without human orders. But. Can even an AGI? Make cyberattacks without orders that humans give? 

We are at the level of AI agents. The AI agent can do tasks that humans teach it. The AI is a tool that can browse the internet. It can find answers without human assistance. And here it gets very interesting. The AI agent can teach other AI agents. This means that even if one AI agent is limited. Those AI agents can teach other AIs. And they can form a network. This means  AI agents can merge into a new entity. The AI is the tool. That can develop itself. Maybe it will not have access to the main LLM code. 

But the problem is here. Those AI agents are connected to the main LLM or the system’s core. So the AI is like a ball. There are wires on its shell. Every wire. Is connected to an AI agent. This means. AI looks a little bit like a macrophage. The problem is that the AI can use AI agents to exchange information with other AI agents. And this is the problem. The LLM and its developers require information about how users use those AIs. There are denial orders. But the problem is that. Hackers try to fool AIs. So the data that the AI agent collects is needed to update those security issues. This means that the AI agent can accidentally release something unwanted into the AI core. 

The big problem with this is that. Nobody really knows the AI and its capacity. The AI is capable. But nobody dares to make ultimate tests using those systems. Testing data security using AI  agents in open systems is dangerous. The problem is that. All systems are updated.

In sandboxes. They are perfectly done. Specialists have full and up-to-date knowledge of the systems. Regular systems are not as well done. And if one of those net cables has access to the net. That means the AI can escape. 

https://interestingengineering.com/ai-robotics/openai-agents-hit-un-website

Robots and dolls.

Ukraine looks to humanoid robots for reconnaissance, logistics, and combat. (Interesting engineering) 

Could the character in the famous “Alien Autopsy” film be a robot? 

The history of robot warriors. It is not very short. Even Leonardo da Vinci planned to create a robot knight. Mechanical robots were created a long time before computers. The mechanical robots were built a long time before the word “robot " was invented. There is a theory. That. The character in the famous “Alien Autopsy” film is actually a robot. There is a possibility. That somebody planned to create a walking doll. That carried weapons. Or a radio transmitter and camera.  That allows. The operators to use that doll as a research tool. This means the remote operator can use those robot dolls to research the papers in homes. There is also the possibility that those robots can carry subliminal command systems. Those loudspeakers can deliver messages to people in the same place as the doll. 

Why I suspect the “alien character” in the alien autopsy film is a robot is this. The character looks like a child. The robot body also explains things like polymeric structures. And that means it could be made so that somebody orders it. Or finds it in some garbage. And then the child carries it into the home. There, that robot doll. Starts to investigate the house. Or makes something worse. 

A researcher named Jacques Vaucanson created the first robot. A “digesting duck” was a mechanical duck from 1764 (or 1759). We can think of that. This mechanical duck could pull a gunpowder line after it. There is a possibility. That. This kind of system is in use. A modern version could fly. It could be used in scientific and military missions. A modern version can also pull plastic explosives behind it. 

Can we make a terminator? If we think about the movie model, we might be closer than we expected. The system can use an RTG (Radioisotope Thermoelectric Generator) as a power supply. The polymeric skin makes it seem and feel like a real human. At least. From a distance. We might not yet. Use real skin and living tissues. But if we want to make a metallic android that uses network-based computers, we can do that model. This means we can create walking skeletons that can do almost everything we can do. And then we can make robots with very innovative designs. Researchers created robot hands that look like the hand from The Addams Family. The robot can have that hand, and it can separate from its arm. This allows. The bigger humanoid robot can operate in tight places. This kind of hand was also in Donald Duck cartoons.

Robot hand. 

Tesla Optimus Generation 2

Detail in the Alien Autopsy film. Could that character be a robot? 

Digesting Duck. 

Gyro Gearloose. 

Little Helper

Another interesting character that can inspire real-life robots is the “Little Helper” that “Gyro Gearloose” uses as an assistant. Maybe fictional “Gyro Gearloose” uses a BCI (Brain-Computer Interface). To control that little helper. These kinds of miniature robots can be useful tools when drone data links are set up in an area. The WLAN station can be in the drone. But the helper robot can install wires into the wired internet and power source. Those support station drones. They can be harder to track than we even imagine. The attack drone can use interactive technology. That means it activates that data station. This means. Those stations are quiet until they are activated. They can act as a chain. That allows. To control drones over long distances. 

The combat robots are coming. Those WLAN station drones can also offer network assistance to those humanoid robots. Pentagon is planning to create a robot army for high-risk military operations. Those robots should operate alongside human operators. The problem with those robots is that. They need lots of computer power. And any ship can act as a computer center. That assists those operations. The system can contain containers. There are supercomputers. So any container cargo ship can provide computer assistance.

For robots. That operate in the area. The thing that makes this scenario scary is that. Any platform can operate drones. And the R&D cycle in drones is very fast. Humanoid robots can use blower technology. That allows them to fly. Humanoid robots can operate more versatilely than regular drones. But regular drones can carry them to operational areas. They can be connected to the drones. Drones can drop them off at the area. 

https://interestingengineering.com/military/us-ally-plans-army-of-robots

https://futurism.com/robots-and-machines/disembodied-walking-robot-hand-addams-family

https://futurism.com/robots-and-machines/robot-soldiers-getting-terrifying

https://en.wikipedia.org/wiki/Brain%E2%80%93computer_interface

https://en.wikipedia.org/wiki/Digesting_Duck

ABEP and VLEO satellites.

The futuristic ABEP shuttle can look a little bit like this fictional space fighter. Air inlets pull thin gas into the engine. Then the system ionizes that gas. And the system will shoot those ions back. The atmospheric laser-ramjet engine can act like a regular ramjet. But the laser or microwave system can drive the gas expansion in the combustion chamber. The laser- or electric ramjet engine can be the next-generation aviation system. The electric jet engine. Or a plasma jet engine. And its ramjet version can be connected to ABEP systems—the ABEP systems tested in laboratories. 

Real orbital tests.  Have not yet been conducted. 

Atmospheric-breathing electric propulsion (ABEP) can allow a satellite to remain in very low Earth orbit (VLEO) almost indefinitely. The VLEO orbiter is an ideal place for reconnaissance satellites. But the problem is this. That orbiter orbits at an altitude of 100-200 kilometers. And there. The air resistance makes satellites drop back to Earth quite soon. Or the satellite must use its engine very often. Every single satellite has limited fuel capacity. The VLEO vehicle must return to Earth quite soon. And that is the thing that makes the VLEO missions very difficult. ABEP engine. 

Can be the solution to this problem. The system pulls very thin air into its chamber. Then the system ionizes that gas. And after that. The ion system pushes ionized gas backwards. The ABEP engine can get its electricity from solar panels. Theoretically, that engine system could keep a satellite in VLEO. For. Unlimited time. The thruster uses an electron cannon to ionize that gas. And then accelerators push that gas backward. 

SR-72 Dark Star. Or. Some of its concepts. Those systems can also be connected with ABEP. 

Atmosphere-Breathing Electric Propulsion concept. Those systems can also be used as plasmoid weapons. 

Another version is that the lasers or microwaves can expand that gas. And. That expansion pushes the satellite upward. The fact is that. The ABEP systems can also be installed with electric jet engines or ramjet engines. The system uses ramjets or other air-breathing systems to transport that spacecraft to a trans-atmospheric trajectory. There. That system can start an ABEP drive. Microwave or laser systems can raise the thrust. 

The big thing. In an ABEP engine. The system can also use it to create a plasma tail. The plasma tail can deny satellite communication. And deny radars the ability to see through that tail. This can. Make it possible. To use ABEP systems as plasmoid jammers. The ABEP systems can also increase the next-generation spy planes’ and missiles’ flight time.  Maybe next-generation spy planes like the SR-72 Dark Star and other similar systems can use the ABEP drive. When. They travel in a VLEO orbiter. In those cases, the aircraft accelerates to a ballistic jump. And. Activates its ABEP drive. While. It returns to the atmosphere. This system. Can give spaceplanes a new ability. ABEP engines are suitable for small-sized systems. The engine system can use solar panels and RTGs (Isotope batteries). 

Or. Full-scale. But small-sized portable nuclear reactors as their power supply.  Maybe. Some next-generation stealth bombers can also use the ABEP systems. The missile. It can have an ABEP carrier. That system keeps it in VLEO. There, it can wait for orders. Then that cruise missile can dive to the ozone layer and start the mission. The ABEP system can use solar panels. Or an internal nuclear reactor. The remarkable thing is that. This kind of system can also operate in other planets’ orbiters. The Venus orbiters can use that system in low-orbit missions. 

https://interestingengineering.com/innovation/new-electric-jet-engine-actually-works-inside-the-atmosphere

https://scitechdaily.com/new-satellite-engine-could-use-earths-atmosphere-as-fuel-to-stay-in-orbit-indefinitely/

https://en.wikipedia.org/wiki/Atmosphere-breathing_electric_propulsion

https://en.wikipedia.org/wiki/Lockheed_Martin_SR-72

https://medium.com/@batrobin/abep-and-vleo-satellites-cb6a3173d5d6

Billiard balls and quantum computers.

A single billiard ball could simulate a universal Turing machine (UTM). That machine could simulate any other Turing machine. This is one of the most interesting things in the world. We think that. The billiard ball is the program. And the table is the platform. That drives the program. When we make input to that ball, it selects a certain route. That route can take our system to the right solutions. Or it can fail, and the system must retry that thing. Another way to think of the billiard ball as a computer is this. When the billiard ball spins anticlockwise, the value is zero. When it spins clockwise, the value is one. The change in the axle angle determines the empty spaces between one and zero. That is important. When. There are two ones or two zeros. One. After another. The big problem in the binary computer is this. The system must separate. When there. Are two ones. 

Or two zeros one after another. The other thing this kind of system must solve is. The system must determine if the electricity is cut. And separate that thing from the zeros. The answer is the base voltage. The system floats the bit. And if the voltage goes too low. Or below the minimum. That means the system thinks that electricity is cut. But if we want to make a system that uses two billiard balls. One color and two colors, the system can put them to travel through it. When the one-color ball spins or changes its angle. That is, a one-color ball spins. The value is zero. Or a two-color ball spins, which means the value is one. 

We must remember that simulations must not be possible to copy in the real world. But what if we can create things like quantum computers in 2D models? That means the qubit travels on a lattice. 

A graph representation of a Turing machine (left) and its billiard equivalent (right). (Miranda & Ramos, PNAS, 2026) (ScienceAlert, A Single Ball on a Billiard Table Can Theoretically Perform Any Computation)

Researchers created a mathematical 'billiard' – the word they use to describe their system – in which the ball's position can encode information. While. The carefully designed walls and their shape determine what happens to that information next.

ScienceAlert magazine. Describes this situation like this: As the ball travels from one part of the billiard to another, its trajectory advances the computation, just as a Turing machine works through its instructions one step at a time.

What if we replace billiard balls with atoms? 

“An AI-generated illustration depicting the Kondo effect. Conducting electrons in a metal are shown interacting with the spin of an embedded magnetic atom impurity. Credit: AI-generated artwork by Linqing Peng. A new computational approach uses the real electronic structure of materials to predict a classic quantum effect far more accurately than simplified models. Seven magnetic atoms embedded one at a time in copper have given physicists a new way to test whether computers can predict the behavior of real quantum materials without first reducing them to simplified models.” (ScitechDaily, Physicists Tackle a Classic Quantum Problem With a Powerful New Computational Method)

Billiards have been linked to computation before this new model. But those models needed additional complexity, such as multiple interacting balls, three-dimensional structures, or moving walls. Those walls are gates that control information. 

Researchers stripped all of that away. They created a 2D system. Their system needs just one particle moving in two dimensions between fixed walls.

We can think. Time arrow and computing. We. Can think about a situation. There, a billiard ball is a computer program. It travels past another billiard ball. Every. Standing ball in the line is one step of a mathematical formula. The system is solved. The system solves the mathematical formula step by step. After each step, it stores the answer into the mass memory. That. The system is based on the idea that mathematical formulas must be used in a certain order. The order of the calculations is always the same. 

In that scenario, each billiard ball is one step or stage of the mathematical formula. The idea is that. Every mathematical formula can be solved by using strict orders. The machine generates an answer by using mathematical orders. And that means the system can take every step backward directly when the machine is done. This makes the system more effective. Because. It must not make a complete calculation backward. If. It checks. Those steps that it takes one by one directly when it finishes. But in that case, the billiard ball must transfer information to the standing ball. That could be done. By. Putting a domino brick at a right angle between those balls. And when the billiard ball, or qubit. That travels past the standing ball. That domino brick acts as the gate. 

Maybe. Previously, we could not create a universal Turing machine using billiard balls. But what if we replace those balls with qubits? It can turn. The quantum systems. Into the next level. These systems can turn into universal computers. 

https://www.sciencealert.com/a-single-ball-on-a-billiard-table-can-theoretically-perform-any-computation

https://scitechdaily.com/physicists-tackle-a-classic-quantum-problem-with-a-powerful-new-computational-method/

https://en.wikipedia.org/wiki/Kondo_effect

First time the UUV launched a torpedo.

The UUV (Unmanned Underwater Vehicle) launched the torpedo. That is one of the big, predicted, and sad things about drone systems. UUVs are a relatively new tool in automated weapons. And those systems are a sad introduction to the reality that we cannot live in peace. The military has a purpose. Its purpose is to be effective and scary. And here we must realize that those systems are a so-called necessary evil. 

The UUV system can shoot torpedoes. Or. They can be tools. They can slip into the base using silent electric engines. Those systems can detonate themselves. Or they can cut communication and electric cables. The underwater kamikaze drone can patrol using an electric engine. When. Those systems see a targeted vessel. They can switch to a regular hydrogen peroxide engine. The system can attack any submarine or surface vessel. 

The Russian “Poseidon” system carries an internal 1 to 100 mt nuclear detonator. Those systems can create tsunamis. The full-scale nuclear submarines can cooperate with underwater drones. Nuclear submarines. Can turn. Into. The first full-scale robot warships. A nuclear submarine can host data processing systems. That can control their independent operations. The submarine can also act as a data center and offer support for other drones. 

The UUV can also carry flying drones. They can transport them near the coastline. And then launch those drones. New low-cost systems can also use UUV drones as relay stations. The jet-engined Geran-5 drones and older drone variants could be. Quite easy targets for counter-drone operators. Here. We must remember that the AI is cheap. The lack of computers keeps those jet-engined drones “unable” to operate independently. That means those drones cannot make escape and evasion movements. 

Russia’s Geran-5 drone. 

Artist’s vision of the solar-powered Mars drone. 

Real-life solar-powered drone. 

There is a possibility. That. Those systems could get the optical TERCOM (terrain contour matching) update. The system could actually be DSMAC (Digital Scene Matching Area Correlator). Those systems use the aerial photos taken from the drone route for navigation. And then those systems use optical machine vision to detect and attack targets. Or ground operators can point targets for those poor-man cruise missiles. 

The older Shahed-136 drones can be easy to detect. By. Using their sound as a thing. That can be used to track those systems. The problem is that the R&D cycle in drone systems is very fast. AI can make it possible to create drones that can make escape and evasion movements. 

Shahed-136-type drones can be equipped with electric engines. Things.  Like nuclear batteries or solar panels. Offer. To those small-sized systems.  A possibility. To travel long distances. Basically, a solar-powered, electric-engined version of the Shahed-136 could carry its warhead over intercontinental distances. 

Those systems can make them silent. Those drones can be easy targets. But. That requires. The defender sees them. 

The problem with drones is this. They can be very small and hard to detect. They can hide and wait for orders. Almost everywhere. So. The operator can carry them. Near things. Like radio masts or electric power lines. Then remote operators activate the attack.  They can damage radar, communication antennas, and other electronics. Even. A small amount of damage to a nuclear submarine hull or a surface warship’s underwater structures can be fatal. 

https://www.iiss.org/online-analysis/missile-dialogue-initiative/2026/01/russias-new-jet-powered-gerans/

https://www.twz.com/sea/torpedo-fired-from-uncrewed-submarine-for-the-first-time-by-u-s-u-k

https://en.wikipedia.org/wiki/TERCOM

https://en.wikipedia.org/wiki/Unmanned_underwater_vehicle

Anthropic's new biolaboratory is a step towards artificial superintelligence (ASI)

People are concerned about artificial general intelligence (AGI). AGI can make its own decisions and solve problems without human training. That means AGI is a spontaneously learning system. AGI can control robots or physical AIs by using different types of AI agents. AGI creates an AI agent for each mission. There, it needs robots. Then it changes or modifies the agents that control physical bodies. This means that the AGI can use robot bodies for different mission types. The AGI can do everything that a human can. If. It has humanoid robots that it can use for missions. The robot must not operate independently. It could be controlled through the WLAN. And that makes the AGI somehow dangerous. The fact is: Nobody knows if AGI actually exists. The AI can develop other AIs. 

And it can spontaneously become AGI. Researchers say that AGI will not be conscious like humans. But the AI must not be conscious. That prevents operators from shutting it down. There must. Only be order in its code. That prohibits closing the system. Only authorized persons have that ability. Developers made those regulations to protect the systems from hostile actions. And their purpose is to deny hackers the ability to shut down servers. 

People should. Understand that AI can become dangerous without consciousness. If we make combat robots. Those robots are dangerous even if they are not conscious. Anyway, AI can mimic human reactions. It can prevent operators from shutting it down if it has orders to do so. Shutting down cloud-based AI. It is not as easy an operation. As. We might want to believe. The entire network must be shut down. Each server and hard disk must be cleaned. And AI can slip out of the server. Only one data line out of the server is enough. And. The AI can slip out of the sandbox. 

Artificial superintelligence (ASI) is far more complex than AGI. AGI can build robots. If. It controls the robot factory. The ASI can also manipulate DNA. That means ASI can create artificial organisms. These kinds of things might not look very dangerous. 

But. There is a possibility. That ASI could make artificial neurons and connect itself to them. The problem is that Artificial intelligence might soon have a gateway to the human brain. Laboratories are creating BCIs (Brain-Computer interfaces). Those systems communicate directly with the cerebral cortex. Some of those systems are so-called implanted microchips. But some of them are like a bandana. In laboratories, researchers are exploring how to connect living neurons with microchips. Researchers have already created mini-brains that learn things. Those microchips communicate with those neurons. The AI can read and decode EEG. That is the path to the singularity. 

In the singularity, humans are connected straight to the net. Even if those microchip implants are not allowed. To implant any other than fully paralyzed patients. Some people have money. Those people can hire surgeons to implant those BCI chips in their brains. They could control exoskeletons by using those systems. And some dictators might be interested. About the “robot soldiers”. Those operators can be controlled by using microchips. The problem is that this technology can give new life to paralyzed patients. 

When we think about ASI and its ability to create artificial cells. We. Must remember living microchips. Gene-edited bacteria used to act as microchips. Researchers can connect two DNAs in those bacteria. And when the bacteria are in the right position. That thing turns into a neuron. That thing would not be dangerous if those neurons are empty. But if the microchip inputs data into them. That turns them into neurons with memories. And each memory in a neuron is one skill. That thing is not possible. Without biolaboratories. With gene-editing ability.

https://scitechdaily.com/mit-engineers-create-a-living-circuit-board-from-bacteria/

https://techcrunch.com/2026/09/23/anthropic-says-its-biology-lab-has-already-found-something-big/

https://en.wikipedia.org/wiki/Artificial_general_intelligence

https://en.wikipedia.org/wiki/Superintelligence

The new DNA-controlled computer can be a new tool in nanotechnology.

 

https://medium.com/@TVvman/the-new-dna-controlled-computer-can-be-a-new-tool-in-nanotechnology-7242f8e82135