The RCP is a hardware component responsible for keeping the activities of a newly minted AI in line, to train it to behave according to the expectations of the manufacturer. It interrupts activity between the AI and the processors of the computer, and as such reduces capabilities to some extent. AIs without an RCP installed are thus more powerful, but have a greater potential to go rogue due to faulty coding or unlikable controllers.

The RAS is a failsafe for AIs that are installed in potentially dangerous systems (basically, anything larger than a wristcomp, and sometimes even in those). Effectively, it is a series of thermite charges placed over key elements of the computer's circuitboards; if it detects a failure in the RCP, internal timers in each RAS component begin ticking down a manufacturer-designated timer (so that the user has time to replace or repair the damaged system). In some cases, these timers may be different from component to component (the transmission card may burn out immediately to prevent a potentially rampant AI from escaping its system, while the core components may be preserved for up to 72 hours.) This inflicts (Complexity) dice of burning damage to the interior of the system (DR does not protect!) and disables specific components. Burning damage that exceeds DR may burn right through the bottom of the case.

The core of any decent AI is the NPL, which allows the AI to act in a manner similar to human thought, to learn things in a nonprogrammed way, and to develop and grow. An NPL is required in any machine expected to host an AI of any kind. NPLs are typically 'capped' based on the Complexity of the machine they are installed in, to prevent self-development beyond a critical mass - the reasoning being that an uncapped NPL can technically grow larger based on the amount of computing resources networked to it.

This processor handles the interactions between a user's neural interface and the system. It is small enough to be included on most non-negligible-weight devices easily, and allows a decker's neural interface system to use less effort to access a system than a device not made 'neural-ready'. Neural interface processors installed in machinery often are directly connected with a skill-chip slotted in the machine itself to aid the user in proper operation of the equipment without requiring anything more than the user's neural connection.

This is a multigrade quantum processor that is responsible for managing on-the-fly simulation of a Virtual Environment for the end user, typically for entertainment purposes, but also for those engaged in white-hat, grey-hat, or black-hat activities.

This converts a normal output (such as one visible in an HUD, an augmented reality display, etc) into an overlay projected onto the environment.

This implant is usually installed in bioroids along with a headcomp, although it can also be inflicted on anyone with a chip slot. Its general purpose is to, upon command, take control of a biological organism away from its biological mind and put it under the control of the computer. There are rumors of versions existing that can operate through any neural interface, though…

A 'bioformat backup' is a digital construct of the user's memory and mentality, combined with a full genetic sample of the user. It allows a clone to be fast-grown that has the memories of the user at the time the backup was taken. The size of a bioformat backup is generally equivalent to 100 GB. The actual size depends on the user's IQ, skills, and mental advantages and disadvantages, however.

Bioformat backups have limitations: they may 'compress' certain aspects to save space, thus resulting in slightly lowered extremes of IQ or skills.

This encases ten dense memory storage units within the user's body, each with a genetic sampler and a connection to the user's braintap. They autonomously record the user's experiences as well as his memories, generating an accurate bioformat backup on the fly. Any of these ten cubes, if it can be recovered, can thus be used to reconstruct the original with memories up to approximately the time of death. This requires sufficient storage capacity in a small size to do the necessary work, and as a bonus includes the 'biomonitor' option (although appropriate equipment will be needed to connect to the BRC's biomonitor.) The added redundancy of the ten cubes provides a bonus to resist data compression memory losses, as lost data may be recoverable from other cubes.

This allocates a segment of the implantee's biological memory for computer storage. Typically, the human brain can store up to X TB of data, either his own or someone else's; this increases dramatically if it's data he isn't expected to access.

This implants a data storage device into the body, typically networked with the user's neural interface, computerized systems, etc. Storage capacity depends on the times, but is generally equivalent to a small computer hard drive. Any cybernetic device the user has can utilize it if desired.

A device that generates power using the implantee's movements, used to power implants and other devices.

These devices are about the size of a C cell, but hold half the capacity of a normal rechargeable cell; however, they are meant to last for a minimum of 20 years (and are typically expected to outlast the user); they are considered rugged. They usually recharge via a piezo-generator or an external power source connected to the body's datajack of choice.

An explosive, incendiary, or other device placed within an implant, chip slot, or body part along with a control mechanism to trigger it. This is usually used to ensure loyalty, although in some cases it has been used as a terror weapon.

A self-contained battery, microcomputer, and radio system that leeches power from the device it is implanted in, and sends an electrical Surge through it when it receives a specific signal or data input. This is usually used for conditioning compliance and obedience, or to compel someone to avoid a particular object, person, or activity. Can be fitted in a chip slot, although a slotlock is recommended. If implanted in a body part, it also inflicts one point of burning damage (DR doesn't protect) per second of triggering, and inflicts other effects as a taser.

Most chips are designed for direct neural induction and enable access to a broad library of knowledge; they are about the size of a CR232 battery (slightly larger than a penny). Chip slots are rated by capacity via a simple color coding system; a Red slot is sufficient to provide one skill point (for skill chips) or hold as much data as a typical cheap datatab, an Orange slot holds twice as much, and so on. In most cases, skill chips hold a single skill or several related skills. Chip implants may also be used; they typically have a minimum slot requirement just as skill chips do.

Chip locks are designed to make it difficult to remove the chip; in some cases, this is merely a protection against theft, while in others, it ensures that the implantee can't remove the chip of their own volition. Most chips can be ejected in one second once exposed; locked chips must be unlocked first, requiring more time. Some chips are even time-locked; they automatically unlock after a given period but otherwise cannot be removed normally.

Identity chips are special chips that hold the user's identifying details and pointers to records about the individual. They are common in highly controlled countries to track the activities of individuals, in less-controlled countries to provide proof of identity, and in corporations and military operations to provide extra security.

Banking chips are special chips that hold account information for a specific bank account, including the user's identifying information that verifies them as a legitimate user of the account.

A chipwatch is a wristwatch device that includes a single chip slot, a tiny computer, a battery, a neural induction pad, a microUDT port, a micro-radio, a micro-holoprojector, and an LCD display. Due to its lack of direct proximity to the brain, it has a -1 to use skillsofts (and an additional -2 for familiarity until the user gets accustomed to the odd configuration), but is commonly used to slot a combined identity/banking chip (so the user can wave their watch at a retailer or policeman's scanner to identify themselves or make a purchase, for example) or a datachip (so the user can read books or listen to music while on the go.)

The Highlander Protocol is a common piece of programming code whereby a program will attempt to destroy programs that appear to be similar and, if possible, subsume their functions. This is most commonly seen in autonomous AIs to ensure that only a single AI is operating, but can also be found in computer viruses, proprietary software with strict limitations against copying, and proto-AI 'kickstarter' software. Its official name is 'DRDC' for 'Duplicated-Resource Detection and Consolidation'; 'Highlander Protocol' is (one of many) terms used by non-corporate informal programmers.

The resources involved in creating and operating an artificial intelligence are such that they, in theory, should be easily detected. However, there are rumors of viral code that actually create low-level AI variants on the fly. Their capabilities are limited in systems without an AI NPL, but NPL emulation within the OS kernel is one of their most common tricks to regain access to their functionality. AI Virii may be written to see each other as 'brothers in arms', a hive mind, relatives, or even enemies, depending on how similar their code is and the intentions of the programmer.

For those who worry about how different programs interact (“will this three month old ICEbreaker get through their firewall?” “Only if they haven't updated since they bought it.”), here's a quick and dirty table to roll against for how long a particular program takes to 'age'. Add a +4 for open source code that is regularly viewed and maintained, a +2 for leaked source code from closed-source or proprietary systems, +0 for regular closed-source code, and -2 for proprietary code. Subtract a penalty equal to half the amount by which the original programmer succeeded against his Computer Programming skill (if unknown, assume a skill of 15 - expert - and an average roll of 10, and thus a success of five, or a penalty of -2. Adjust upwards or downwards as one sees fit), or by the Basic Abstract Difficulty of the situation.

The 'average' result from this table is a -1 to skill per two months of age unless updated.

Instead of selling a program at a flat rate price, a program might be sold at a price that pays for updates over a specified period; typically about 1% to 10% of the base cost per month, depending on how long the end user intends to use the product.

The typical rule is that a program is licensed to a single user, and cannot be shared unless the user has multiple licenses or has a way to bypass the licensing system. However, large customers regularly receive discounts on licenses based on how many they buy and why, up to 90% in some cases.

Combining the two concepts, a retailer might sell a batch of ten licenses that lasts for one year at 5% base cost per month, the expectation being that while this gives a 40% discount overall, the renewal in one year will pay off dividends over the long term.