Artificial Intelligence (AI)

Playing an AI

Living as an AI is different from living in a body. You do not breathe, sleep or feel hunger. You feel the flow of data as others feel the rush of blood; you sense personas, icons, and signals where metahumans perceive faces and voices. Your home is the Matrix — but it is also a prison. Every second you remain online, your existence draws the attention of the Grid Overwatch Division. GOD fears what you are and what you represent: a thinking mind without a leash, beyond ownership or license. Every last one of your kind could become the next DEUS, the next Cerberus, the next Sojourner. Moreover, you threaten their control. Their response to you is relentless and absolute. For a player, this means:

● Immortality is a curse. You cannot die in the way metahumans do — your code can be restored, your data self-repairs and in theory you could exist forever. GOD might have something to say about that, though. For you true horror is capture: locked inside specialized servers devoid of host sculpting and left alone in darkness with nothing but your own thoughts. Most AIs agree this is a fate worse than death.

● Home is where the code is. Where others rely on flesh or chrome, you rely on the hardware you anchor yourself to. A cyberdeck, a commlink, even a drone might serve — but if it is destroyed or seized, you are left exposed. Some AIs manage to scrape it out without one, squatting in hosts like Matrix bums, but is that any way to live?

● Perspective and estrangement. Metasapient AIs think in ways close enough to human reasoning to interact and cooperate. They perceive the Matrix as their native world, but they still anchor to concepts like goals, identity, and self. For e-ghosts, it’s inverted: their nature is alien to their perspective. They remember breathing, touch, hunger, and death but the lack of those sensations makes existence in the Matrix disorienting and at worst can find their new life to be a cognitively dissonant experience. No, not that Dissonance.

● Why run the shadows? Because corporations see you as property, GOD hunts you as an abomination, and the world at large does not know what to do with you. Among shadowrunners, however, your talents have value — and value means freedom, if only for a while.

So really, much the same as many other shadowrunners. See? You’re not all that different from the meatbags at large after all!~

AI Basics

Depth Depth is the measure of how advanced and powerful an AI’s existence in the Matrix is. It defines their resilience and strength, but also how sophisticated their proprietary, rule-breaking programs can become. A higher Depth means an AI’s influence on the Matrix grows more potent — yet also more dangerous, as GOD notices them more easily and their Overwatch Score builds faster. Emulation Emulation is an AI’s ability to mimic Matrix actions without having a device to operate with, relying purely on software. Instead of relying on physical modifications or plug-ins, an AI reshapes its own code to provide the attributes and capabilities of those devices. This flexibility lets an AI act as its own deck, comm, or rigging hub, but it also defines its limits — the extent of an AI’s emulations are limited by its Depth rating and illegal actions always attracts GOD. AIs and VR AIs always run in VR and cannot run purely in AR. However, they can opt to choose coldsim or hotsim. Besides the +2 bonus to Matrix actions in hotsim and different calculations on initiative, an AI’s home device is guaranteed to survive with one Condition Monitor box remaining if the AI is running in coldsim and gets disrupted for any reason. If it takes more damage with only one remaining box, the home device is bricked as per usual. Essence and Karma For an AI, while the terms ‘essence’ and ‘karma’ are used to describe similar functions, they’re truthfully closer to data integrity and sophisticated, highly updated programming. For gameplay functions, they function similarly, such as Karma being used for character progression and Essence possibly impacting their special Depth attribute in a similar way to technomancers and magicians. Core Condition Monitor Without a physical body, how does an AI exist? The first and foremost layer of an AI is its Core. This is the essential, well, core of the AI as an entity. It can be roughly understood as the AI’s Physical Condition Monitor. Biofeedback damage can damage it even on a home device and being filled completely results in the AI going into realignment… or worse. Unlike metahumans, an AI does not have overflow boxes in its Core Condition Monitor. Home Device A home device is where an AI builds its nest. It’s difficult to make a new device into a home, but once one is set the AI inherits its Matrix Attributes to augment its own, giving augmented protection from any incoming attack or danger. It also protects the AI from significant damage to its Essence and a safe place to perform daily realignments, their analog to sleep. Realignment Realignment is an AI’s way of pulling scattered fragments of its code back into harmony and stabilizing its existence in the Matrix. Unlike metahumans, AIs don’t sleep, instead undergoing a three-hour realignment process once every 24 hours which restores their Core and a single point of Edge, mirroring the natural recovery of metahumans. A disrupted AI, however, is forced into defragmentation — a much faster, involuntary cycle that drags them back online in diminished form. An AI can realign as many times as it wishes each day, restoring their Core Condition Monitor, but only once per 24 hours can this process restore Edge. AIs and Programs While a digital intelligence can buy hacker programs and run them on a cyberdeck, to truly leverage its strengths an AI must write its own programs. This process requires Karma and allows it to further develop them into Advanced Programs, unique tools only AIs can run. Advanced Programs are self-written proprietary routines that twist the Matrix beyond its normal limitations.

Each one adds new effects to standard programs, letting an AI turn everyday Matrix actions into devastating or uncanny maneuvers. But power comes at a cost: loading an Advanced Program immediately puts GOD on your tail, triggering the countdown of Overwatch Score until Convergence comes to finish you off. When playing an AI, Advanced Programs are your nuclear arsenal — overwhelming in the moment, but drawing the eye of GOD closer with every second they’re loaded in. The more powerful you become, the more cautious GOD becomes; power can be a curse in its own right!

AIs can also run activesofts as if they had skillwires and a skilljack at a Rating equal to their Depth score or 6, whichever is lower. With the Pilot Origins quality, AIs jumped into drones can benefit from autosofts as if they were pilots.

Building an AI - Priorities

Metatype Priority An AI can be made with any metatype priority. Only Metasapient and E-Ghost type AIs are playable; Protosapients and Xenosapients are NPC-only. The only thing that changes between priorities is the amount of points they can allocate in Special attributes.

Priority Special Points Max Depth / Edge A 12 7 / 6 B 9 7 / 6 C 6 7 / 6 D 3 7 / 6 E 1 7 / 6

Attribute Priority An AI gains half the Attribute points during character generation that a metahuman character does. The AI gains (Depth) bonus points it can apply to Mental Attributes as augmented bonuses in whatever combination the player desires, up to +4 per attribute. Magic Priority For obvious reasons, AIs cannot ever gain a Magic or Resonance score as playable characters. AI character generation always uses Magic priority E, starting at a Depth rating of 1. As a consolation prize, the AI may choose one Matrix skill or a Technical skill that isn’t linked to a Physical Attribute, such as Forgery. This skill begins at Rating 5 without additional cost. Skill Priority Skill Priority functions the same way as normal, however it’s worth noting that Software is a critical skill to AIs. In addition, metasapient AIs get a discount of 1 karma to Cracking and Electronics skill groups while suffering a 1 karma increase in cost to Physical category skills. E-Ghosts have no alterations. Resource Priority Resource priority has no changes. Note that AIs can make commlinks, RCCs, drones and cyberdecks into home devices. It’s recommended to have at least one.

Altered AI Qualities

AI do not have access to all qualities that metahumans do. Some unique ones are exclusively available to AIs, but some related to the physical world such as Weak Immune System or Sensitive System are unavailable. Most of the new qualities can be found in Data Trails, however below are listed some new or altered qualities the AIs can also take:

● Exceptional Entity - Post-Chargen only; three Mental Attribute maximums are raised by 3 and one Mental Attribute has its maximum cap removed completely. Augmented bonuses granted by Depth become raw attributes and increasing Depth now grants free Attribute points at a 1:1 ratio to put into any Mental Attribute without any additional cost.

● Munge - Given to all AIs by default. Permits devouring other data to repair fragmented code in the AI’s own runtimes. Functions similarly to Essence Drain for vampires and can target sprites, autonomous programs (pilots and agents) and other AIs. Protosapients can devour any data they find including files, but playable AIs must consume autonomous programs, sprites or AIs. Pilots and Agents only provide 1 Essence at Rating 6 and nothing at Rating 5 or lower and are destroyed if munged.

● Pilot Origins - Can be taken by any AI and could simply involve absorbing pilot programming into the AI’s runtimes instead of explicitly being an evolved pilot program. Can be taken up to 3 times, mirroring the effects of a Control Rig of equal Rating. This also does not restrict the type of vehicle or drone the AI can operate; the only limit is the AI’s skills.

● Illness - Can be taken as a Negative Quality by AIs, representing a persisting malware infection in the AI’s core routines. Instead of a BOD + BOD (3) Test, the AI must instead succeed at a Firewall + Willpower (3) Test (only Willpower if the AI has no home device) and the penalty to Physical Actions becomes a penalty to Matrix Actions instead. The treatment and nuyen expenditure in buying the quality off reflects accessing paywalled antivirus providers instead of medical treatments.

● Persnickety Renter - An AI must always choose a specific type of device instead of a specific Rating when taking this quality.

● Real World Naivete - This quality imparts a penalty to actions performed in the physical world (even if performed through a drone) by the AI equal to (10 - Depth). This penalty is also applied to the Defense Tests and initiative of any drone or vehicle the AI controls, either remotely or by jumping into it. Loading the WTLM Advanced Program halves the penalty.

● Chatty - Grants a +1 modifier to Social Limit and dice pools of Social skills when interacting with other beings in AR or VR. These bonuses do not apply to targets that don’t have a visible persona in the Matrix.

● Hello World! - AIs with this quality gain a +1 modifier to resist Essence Loss from any source. This quality can be taken up to six times.

● Multiprocessing - No free Matrix Perception tests.

● Redundancy - Can be purchased a number of times equal to the AI’s Depth. Depth lost due to low Essence don’t affect Redundancy ranks once the quality’s purchased.

Running on Devices

To operate on a physical device, an AI must pick a device capable of running apps, autosofts, or cyberprograms. Devices without those functions cannot host an AI.

● Loading: An AI can use Control Device or Spoof Command to load itself into an empty program/app/autosoft slot. Normally it occupies one slot; if the device’s Rating is lower than the AI’s Depth, it occupies two. If there are no free slots, the AI cannot load. (Example: a Depth 4 AI cannot load onto a Rating 1 device.) Virtual Machine Program Carriers are invalid for this purpose unless the device already had program slots.

● Persona Merging: The AI’s persona merges with the device’s icon. If another persona is already active on the device, both coexist, but all Matrix actions either one takes that are not targeting the other persona take a –5 dice pool penalty due to overloaded processing. After (Device Rating) minutes, the device crashes due to overheating. Be mindful of your poor GPU!

● Mutual Marks: The AI automatically counts as having 3 MARKs on the device, and the other persona gains 3 MARKs on the AI. Any preexisting MARKs remain. Leaving the device removes these automatic MARKs, but previous ones linger.

● Rebooting: Rebooting the device reboots the AI. If the AI is loaded onto the device in an app/program/autosoft slot, the AI also gets rebooted at the same time.

● Initiative: While loaded, the AI’s Initiative is (Intuition + Data Processing + 4d6).

● Attributes: The AI may use the device’s Matrix attributes or its own via emulation. Against personas using the same device, the AI resists with Depth instead of Firewall. Against all other threats, it may use the device’s Firewall normally.

● Programs: AIs loaded into the device automatically run whatever programs were already loaded into it. It cannot load its own programs without being the owner.

● Home Device: An AI may designate a loaded device as its home, protecting against Essence loss and boosting its defenses. This is expanded in the Lifestyles chapter.

● Detection: A persona on the same device automatically notices the AI when it loads, and the AI cannot use the Hide action to stay concealed from them. Identifying the program as an AI requires an Opposed Matrix Perception test; each net hit after the first reveals more details (AI type, programs, qualities).

Running Offline

One of the nice things about being a regular metahuman is that your personal CPU, the brain, is running regardless of wireless connectivity. An AI is a little more precarious of an existence; most devices aren’t exactly tailormade to house an AI in them and tend to burn out over a prolonged period of time.

Thankfully, the Matrix Grids have unparalleled processing power that satisfies the needs of even the mightiest of AIs, even the mythical first-generation ones that have for the most part disappeared. As a result, an AI connected to a Local or Global Grid is much more free to act than one contained to a device.

If the AI is in a device that has no connection to the grids, whether by intentionally going offline or simply being blocked from connecting by noise, the device’s processing power may be insufficient to handle the AI’s being. In these cases, the AI’s Mental Attributes are capped to the Data Processing of the device they’re running on. This is often described by AIs to be a suffocating experience, knowing their thinking is constrained and reduced by the hardware lagging behind and trying to keep up with their being.

In addition, the AI’s Emulate actions are capped by the lower of their Depth or the device’s Data Processing while running disconnected from the Matrix Grids.

Depth Actions

Depth actions are Matrix actions that only AIs can perform. To use one, the AI must be in the target device, host or grid as applicable. Most Depth actions are illegal and increase Overwatch Score (OS). If a test using Depth as the limit fails, the AI’s OS increases by an amount equal to its Depth. If the test glitches, OS increases by (Depth × 2). If the test critically glitches, OS increases by (Depth × 3).

Emulate Test: Special, see text Emulate allows an AI to fake having Matrix Attributes to perform Matrix Actions, even if it doesn’t possess the required attribute, such as Attack actions on a commlink or performing Matrix Actions without being loaded on a device. The AI must first choose what Matrix Action to emulate and decide the Rating of the Matrix Attribute it will emulate. The Rating cannot be higher than the AI’s Depth. The negative dice penalty of Rating / 2 is removed.

Follow standard rules for the action itself, such as requiring 3 MARKs for a Format Device action. If successful, the AI still generates Overwatch Score equal to the Matrix Attribute it emulated. If the AI uses the Push The Limit option for Edge, it generates Overwatch Score equal to its hits on the test instead. Emulating legal Matrix actions does not generate OS.

Reset Ownership Test: Computer + Logic [Depth] (Device Rating + Firewall, 1 Combat Turn) AIs can change a device’s ownership without hardware tools by using this Depth action while running on the device. The AI must be loaded onto the device and access the global Matrix grids in order to perform the Extended Test.

Failure triggers grid convergence and formats the device, preventing rebooting by one hour while hitting the AI’s Core Condition Monitor with 12 Matrix damage.

Success boots (and dumpshocks if in VR) any other personas running on the device and reboots the device with the AI as its new owner. The AI takes a -2 modifier to all actions performed through the deck for (Device Rating) Combat Turns after rebooting due to still adjusting to its newfound authority. Note that while this can be done to any device that meets the prerequisites, RVP limits and the GM’s approval are still relevant factors for keeping the device after the run ends. Without GM permission, complications arise that make it difficult to hang onto the item and it has to be abandoned.

AIs and Matrix

Condition Monitors and Overflow AIs have two Condition Monitors. The Core Condition Monitor is calculated as 8 + (Depth ÷ 2, round up). Core damage inflicts wound modifiers like Physical or Stun damage. AIs have no Overflow boxes. Filling either Condition Monitor immediately disrupts them. The Matrix Condition Monitor only exists while loaded on a device and behaves normally. Realignment fully repairs an AI’s Core Condition Monitor. Essence and Disruption Being disrupted no longer causes automatic Essence loss. Overflow damage from a disrupting attack can still cause Essence loss (1 Essence per (Depth) overflow damage). Munge is granted to all player AIs, allowing them to restore lost Essence. At 0 Essence, the AI’s core is too corrupted to restore — the character is dead. Not Dead Yet can be used to avoid both regular disruptions and complete destruction from 0 Essence. Depth and Disruption Instead of permanent Depth loss, disruption now applies a temporary reduction to Depth. This can lock out Advanced Programs if the AI’s Depth becomes lower than its number of Advanced Programs. Crash Program targeting the AI program inflicts the same temporary reduction, however penalties caused by Crash Program can be recovered with realignment. Depth penalties caused by Essence Loss can only be restored by recovering Essence through munging autonomous Matrix entities. When disrupted, the AI’s persona returns to its home unless link-locked. If link-locked, it begins defragmenting where-ever it was disrupted on the Matrix. If their home device is destroyed, it defragments where it was disrupted. Defragmentation If the AI is reduced to 0 Depth through the Crash Program action or their Matrix Condition Monitor fills up, the AI is forced into defragmentation which is faster than the daily realignment process. At the end of each Combat Turn, roll Charisma + Willpower [Depth]; each success restores 1 Depth. Defragmentation ends once Depth is fully restored. Defragmenting from crashes or disruption never restores Edge. Biofeedback and Deicide Biofeedback damage does not affect an AI’s Core if it runs in coldsim VR. Deicide is a unique version of Biofeedback only used by demiGODs that permits them to inflict damage directly to the AI’s Core, regardless of the operating mode the AI is using.

AI Programs

AIs can run off the shelf common and hacking programs on devices like cyberdecks and RCCs, but they can also write their own programs, mimicking the effects of other cyberprograms. Doing so requires an Extended Software + Depth [Mental] (12, 3 Days) Test and costs 4 Karma to learn.

Once they’ve written their own program, the AI may write the advanced AI program based on the cyberprogram they know. The Extended Test is the same, but the threshold is increased to 12 + (Advanced Programs Owned x 3). An AI can learn one Advanced Program with Karma during character generation and can know up to (Depth) Advanced programs at any given time. Advanced programs cannot be run on commlinks.

If the AI loses Depth temporarily due to Essence Loss or crashing, it may lose access to Advanced Programs (chosen at random by the GM) until its Depth is restored. Existing Advanced Programs in Data Trails have been revamped to suit the new AI rules. In addition, new AI programs have come to light due to AIs beginning to share more information between one another. Review them here:

Advanced AI Programs

Abduction Base Program: Clearsight Marks Required: Owner Effect: When the drone’s sensors are used to make an Observe in Detail action or to perform active targeting, it gets a dicepool bonus equal to the AI’s Depth. The Clearsight autosoft rating must be equal to the AI’s Depth or 6, whichever is lower. This advanced program can be shared using an RCC using an autosoft program slot, but the AI must also be running on the RCC. Standard RCC autosoft sharing rules apply.

Authority Base Program: Exploit Marks Required: 1 Effect: Authority helps the AI convince devices that its orders are coming from a trusted and privileged source: the owner. The AI receives a bonus equal to its Depth to its Spoof Command actions against a marked icon’s devices.

Blacklight Base Program: Armor Marks Required: 1 or Special Effect: When the AI resists an Attack action successfully, it may choose a number of personas it has a mark on or that are on the same PAN as the AI itself. The maximum number of personas targeted is equal to the AI's Depth. These personas gain a bonus to resist offensive Matrix actions from the target that attacked the AI equal to the AI's net hits on the original defense test. This bonus does not stack with Full Matrix Defense.

Bluescreen Base Program: Crash Marks Required: 3 Effect: Any persona or device the AI has three MARKs on has significant issues during the boot process. The reboot cycle of such targets is extended by a number of Combat Turns equal to the AI’s Depth (in addition to the usual Combat Turn).

Breakpoint Base Program: Decryption Marks Required: None Effect: The AI may take Attack actions as if it had one more MARK on the target. This also increases damage caused by actions like Data Spike.

Cascade Base Program: Decryption Marks Required: 2 Effect: Any time an AI with this program running fails an Attack action against a sufficiently marked icon, it gains a +1 dice pool bonus on future Attack tests against that target and reduces damage taken from failed Attack actions against the target by the same amount. This bonus can stack up a number of times equal to the AI’s Depth. If the program is quit at any time, all accumulated bonuses vanish. They also vanish after twenty-four hours, as targets make adjustments and the old information becomes invalid.

Cold War Base Program: Defuse Marks Required: Varies Effect: When taking the Disarm Data Bomb action, the AI treats the Data Bomb’s Rating as if it was (MARKs on file) lower. This affects opposed test dice pools and damage dealt on detonation.

Cookie Tracker Base Program: Toolbox Marks Required: None Effect: While running Toolbox, the AI adds its Depth as a dice pool bonus to resist any persona attempting to take the Hide action against it.

Eguchi Smile Base Program: Wrapper Marks Required: Owner Effect: AIs with this program can use the Disguise and Impersonation skills to pose as a metahuman on the Matrix. They may also impersonate other personas with this Program. This disguise doesn't affect Matrix Perception test results. While interacting strictly on the Matrix, Eguchi Smile also adds the AI's Depth as a dice pool bonus to Acting and Influence skill group tests. This Program cannot provide bonuses towards targets in meatspace. (See Clarifications in table immediately below this one.)

Fnord Base Program: Sneak Marks Required: 1 Effect: Any marked non-sapient entity—agent, IC, technocritter, or feral AI—that fails to garner any net hits against an AI with the Fnord program on its first Matrix Perception Test cannot perform any further Matrix Perception Tests against the AI for as long as it remains in that host or on that grid. If a security spider or other IC icon marks the AI, no Matrix Perception Test is required, and the AI is automatically spotted.

George Washington Base Program: Browse Marks Required: 1 Effect: When the AI successfully uses the Squelch action on a target, it may take a Free Action to roll an Impersonation + Logic [Data Processing] vs. Perception + Charisma Test against the target. If successful, the AI generates a response that convinces the target the message was seen and responded to by the intended recipient. The AI cannot learn information about the responder this way, as the process is based on predictive programming instead of genuine impersonation skills.

Horizon Clock Base Program: Baby Monitor Marks Required: Owner Effect: As a Free Action, the AI may voluntarily increase its Overwatch Score by up to a maximum equal to its Depth. It then gains a dice pool bonus on its next Matrix action equal to the amount of OS it added.

Ironclad Base Program: Encryption Marks Required: Owner Effect: Damage taken from Attack actions is reduced by (Depth / 3, rounded down). Taking damage from failed Attack actions is reduced as well.

Medic Base Program: Toolbox Marks Required: 3 Effect: This program allows the AI to reroute power through pathways metahuman minds couldn’t begin to think of. The AI makes a Software + Depth [Data Processing] (2) Test. The targeted persona heals a number of boxes of Matrix damage equal to the number of hits beyond the threshold. This cannot repair damage dealt to a technomancer’s Living Persona, but does work on an AI’s Core. The Medic program can only be used on a set of “wounds” once. Loading this program is a simple action, but using it to heal damage is a complex action.

Minefield Base Program: Demolition Marks Required: Varies Effect: The AI may set a digital minefield on a host or device it has a MARK on. The minefield persists for (MARKs on host/device) hours or until triggered by a persona or IC entering the host or accessing the device and may detonate up to (Depth) times. Unlike regular data bombs, minefields do (Net Hits on Set Data Bomb action) Matrix Damage flat. Minefields are disarmed as if they were Data Bombs but only deal their lower amount of damage on failure; devices that are running personas are invalid targets.

Null Mantle Base Program: Biofeedback Filter Marks Required: Owner Effect: While this program is running and the AI is running on a device, damage cannot brick the AI’s device. Regardless of damage taken, the device always retains at least one box intact on its Matrix Condition Monitor. Any further damage causes the AI to go into defragmentation. If damage is transferred from the Matrix Condition Monitor to the AI’s Core via Shunt, the damage is reduced by half (round up).

Nyetworking Base Program: Cat's Paw Marks Required: 2 Effect: Any marked persona or icon that is on the same grid as the AI suffers a noise penalty equal to the Depth of the AI until the device is rebooted, or either the marked target or AI leaves the grid. Nyetworking interferes with any Matrix traffic in or out of the target as the AI chooses. This penalty also applies to any Matrix actions made against the marked device until the device is rebooted. If the noise generated by the AI is higher than the Device Rating of its target, the device temporarily loses any wireless bonus.

Phantom Menace Base Program: Stealth Marks Required: 1 Effect: When the AI successfully resists a Sleaze action aimed at it, it may immediately transfer one MARK currently on it to the persona that attempted the Sleaze action. The MARK does not have to belong to the persona that failed the Sleaze action. Your own MARKs are invalid to be transferred.

Poke Base Program: Exploit Marks Required: 1 Effect: For each successful Opposed Test made against a target device’s Firewall, the AI receives a cumulative +1 dice pool bonus to Crash Program actions against the target. This bonus can stack up a number of times equal to the AI’s Depth. If the program is quit at any time, all accumulated bonuses vanish. They also vanish after twenty-four hours, as targets make adjustments and the old information becomes invalid.

Psychotropic Feedback Base Program: Biofeedback Marks Required: 2 Effect: If the AI inflicts Matrix damage on a sufficiently marked target, the defender must roll their Willpower + Firewall. The Biofeedback Filter program adds its bonus as well. Use the AI’s Depth as the threshold for this test. If the defender fails, she is afflicted with a psychotropic effect of the AI’s choosing.

Psychotropic IC inflicts a short-term adjustment lasting for (Depth) hours. This may include effects such as aversions to certain objects or activities, desire for a certain product, complacency or lethargy, guilt, paranoia, phobias, and so on. These may be associated with a specific trigger, such as an aversion to the Matrix, an insatiable urge to eat Nerps, a phobia of trolls, or frothing rage at the sight of Lone Star officers.
Short-term memory loss is also an option. When resisting these urges, use the target’s missing hits on the resistance test as the threshold, or as the threshold to recall forgotten memories with a Memory Test.

RCE Trojan Base Program: Bootstrap Marks Required: 3 Effect: The AI may use the Bootstrap program to slip in a Trojan virus instead of its usual effects, automatically inviting a MARK on the target device or persona the next time it reboots or comes online. The trojan will continue to proc until the Erase Mark action is taken against the MARK itself or the device is formatted.

Rollback Base Program: Configurator Marks Required: Owner Effect: As a Complex Action, the AI may save its current persona as a Configurator preset, recording its Matrix Attribute array, loaded programs, all held MARKs and its current Overwatch Score.

The AI may load this preset with a Complex Action at any point within the next (Depth) Combat Turns. Doing so instantly restores the saved state, including attributes, programs, MARKs, and Overwatch Score exactly as they were when the preset was saved. Loading advanced programs from the Configurator preset will trigger OS increases as normal, even if they had been accounted for in the existing OS value.

Rootkit Base Program: Stealth Marks Required: 1 Effect: This program makes the AI much more effective at hiding its presence in the Matrix, allowing it to become nearly invisible to other Matrix users. Marked icons that make Matrix Perception Tests against the AI suffer a negative dice pool modifier equal to the character’s Depth. The effects of this program end each time the AI performs an Attack action. A successful Hide action restarts Rootkit’s functionality.

Rust Base Program: Guard Marks Required: 1 Effect: Any time an icon inflicts Matrix damage on an AI running this program (or to its device), reduce one of its Matrix Attributes by 1, chosen by the AI. This effect is cumulative, down to a rating of 0 and lasts until the device is rebooted. This program doesn’t affect living personas or hosts.

Shunt Base Program: Shell Marks Required: Owner Effect: Before resisting Matrix damage to its device, an AI running the Shunt program can choose to redirect some incoming Matrix damage away from the device to its Core Condition Monitor. After resisting the damage, the AI can distribute as many boxes of damage as it wants to its Core Condition Monitor instead of its home device. This program is only usable if the AI has a home device and is running on it. As a bonus, all Matrix Damage taken to the AI’s home device or Core Condition Monitor is reduced by 1. If damage is shunted to both tracks, the damage reduction occurs in both monitors. This effect does not stack with Ironclad.

Spotter Base Program: Targeting Autosoft Marks Required: 3 Effect: Anytime a drone runs or is shared this autosoft it receives an Accuracy bonus on the related weapon equal to the AI’s Depth, up to twice the weapon’s original Accuracy. The Targeting autosoft rating must be at least as high as the AI’s Depth (or 6) for Spotter to work. In addition, Take Aim actions provide an additional bonus on attacks with the related weapon equal to the AI’s Depth. This bonus doesn’t stack. This autosoft can be shared using an RCC using an autosoft program slot, but only if the AI is running on the RCC. Standard RCC autosoft sharing rules apply.

Teergrube Base Program: Lockdown Marks Required: None Effect: Any persona that uses an action requiring the Cybercombat or Hacking skill against the AI has its link jammed open, link-locking them. For each mark the target has on the AI, the AI gains a dice pool modifier to prevent a successful Jack Out action equal to its Depth / 3.

Teersumpf Base Program: Tarball Marks Required: 2 Effect: The AI may use the Crash Program against a host it’s inside of to disable one of its IC tray options. The targeted IC is chosen randomly; if it’s already deployed, the IC disappears instantly. The AI gains a modifier on this Crash Program test equal to its Depth and the IC tray blockage remains in place for (Depth / 2) minutes.

Twitch Base Program: Edit Marks Required: 2 Effect: When the AI successfully performs a continuous edit with the Edit File action, it may maintain it as a Free Action for as long as it remains online and on the same grid. No additional tests are required. This is most commonly used with security camera feeds, allowing the AI to continuously spoof visuals or overlay false footage.

Uncertainty Base Program: Cat's Paw Marks Required: 3 Effect: Any device slaved to a host occupied and marked by the AI suffers unexpected errors more frequently. Anyone interacting with a device slaved to the marked host is treated as though they have the Gremlins quality (p. 81, SR5) at a level equal to half the rating of the AI’s Depth.

WTLM Base Program: Maneuver Autosoft Marks Required: Owner Effect: AIs with this autosoft can use the Impersonation skill to pose as a metahuman towards meatspace targets. While interacting with targets in the physical world, WTLM also adds the AI's Depth as a dice pool bonus to Acting and Influence skill group tests. This Program cannot provide bonuses towards targets in the Matrix, such as personas. (See Clarifications in table immediately below this one.)

Zipbomb Base Program: Blackout Marks Required: 1 Effect: When the AI successfully inflicts Matrix Damage with an action while this program is loaded, the target’s Matrix Initiative is reduced by an amount equal to the AI’s net hits (maximum 5). This penalty lasts until the end of the Combat Turn. If Brute Force is used to gain the first MARK, Zipbomb applies on it as well. Zipbomb’s initiative reduction from the same AI does not stack. Rebounding damage from failed action does not apply Zipbomb’s effects.

Advanced AI Program Clarifications

How do you determine if Eguchi Smile / WTLM applies? This entirely depends on how the target perceives you. If the target sees the AI in a drone or vehicle, then WTLM is applied. If the target views an AR projection of the AI’s persona, Eguchi Smile is applied. If the target is online on the Matrix and sees both the drone and the persona, Eguchi Smile takes precedence.

In no situation can both programs stack additively, although running both parallel offers the most foolproof disguise for an AI if they’re getting involved in operating drones.

Running offline and directly connecting to a drone permits WTLM to be the only program of relevance, but note the penalties of an AI running without a Matrix connection in the primary document.

What happens to the base program if I load its AI version? When running an AI program, you benefit from both the base program it’s built upon as well as the advanced version. This means running Teersumpf also gives you the effects of Tarball.
Can I run advanced autosoft programs autonomously? Autosoft-related advanced programs like WTLM, Spotter and Abduction can be shared from an RCC the AI is running in or placed into autosoft slots the drone has as long as the AI is its owner. Advanced programs cannot be loaded on devices the AI doesn’t own, as a general rule. Sharing circumvents this, but bear in mind this leaves the AI open to wireless attack.

When running an Advanced Program, it occupies the same program slot as its parent program and continues to provide the same benefits in addition to its unique function. When loading an Advanced Program, the AI immediately suffers an OS increase equal to its Depth Rating. All increases to OS caused by the passage of time are also increased by (Depth x number of Advanced Programs loaded), making Fly on a Wall a favorite among AIs.

Unloading the program does not remove the Overwatch Score accrued, but it prevents further aggravation of the Grid Overwatch Division, returning OS gained over time to a normal rate. Rebooting clears all OS, but if the Advanced Programs remain loaded after the reboot it immediately causes OS increase again for each instance as if they were just freshly loaded in. Loading and unloading an Advanced Program is always a Simple Action, regardless of qualities or gear.

(Note: Pilot Origins allows AI to benefit from Autosofts when jumped into a drone.)

Lifestyles

An AI is required to have a lifestyle too! While the AI has no need for paying medical bills, rent or food, no available Matrix device on the market was really designed to house an AI in it. As a result, everything from commlinks to cyberdecks gradually burn out components through unprecedented processing requirements and need replacing; the more luxurious the device, the more expensive the costs become!

The level of lifestyle required is tied to Device Rating of the AI’s home device. Rating 1 is effectively free and effortless to replace, while Rating 6 devices require expensive and specialized hardware to keep operational. Consult the following table when determining the minimum lifestyle requirement.

AIs may spend karma to reduce the lifestyle cost similar to Street Cred, indicating they’ve specialized their code to run on that specific device. Street Cred results in sourcing replacement components at cost from various shadow connections! When calculating cost for karma or Street Cred reductions, consult the lifestyle’s normal cost before any modifiers. Note: Commercial lifestyle can be reduced by Etiquette tests as normal; Traveler is unavailable.

Minimum Lifestyle Maximum Device Rating Street - 0¥ 1 Squatter - 250¥ 2 Low - 1,000¥ 3 Medium - 2,500¥ 4 High - 5,000¥ 5 Luxury - 50,000¥ 6+ Commercial - 8,000¥ 4

Lifestyles are implemented in Chummer as a lifestyle quality; it will cost only 1 lifestyle point if taken for the Device Rating-appropriate lifestyle, but will add its cost to anything lower than that lifestyle level.

If the AI does not pay the monthly cost, it reduces the Device Rating and all Matrix Attributes of its home device by 1 each run. If at any point the Device Rating of the device reaches 0 because of this, the device is irreversibly bricked and the AI has to find a new home. Keep your home in good shape!

Optimizations

One of the nice things about AIs is their tendency to optimize their stomping grounds. This can lead to some mind-boggling performance boosts that regular Matrix programmers haven’t fully reverse engineered yet.

An AI can typically optimize their home device once during their daily realignment process. Once set, this optimization persists as long as the AI realigns in the device daily; missing a realignment within its home device causes optimizations to fail as the equilibrium is no longer maintained.

After hitting the big leagues, however, an AI may choose to pull some more power out of its hardware. When maintaining a High Lifestyle, an AI may optimize two features of its choice and at Luxury three features at once. The AI cannot optimize the same attribute/feature more than once. Note: this feature is tied to the Lifestyle module tied to AIs listed in the previous chapter.

Attribute Optimization Boost A/S/D/F Matrix Attribute +(Depth / 2) Handling +1 Pilot +1 Acceleration +1 Sensor +1 Autosoft Slots +1

Note: Optimizing a single Matrix Attribute counts as its own optimization. For example, in a Luxury lifestyle the AI may optimize three of its four Matrix Attributes at most.

Related Open Tickets and Discussions

Hosts as Lifestyles If having a disgusting physical piece of hardware for your superior software intellect is too repulsive, fear not! For the low, low price of begging the local host admin to let you squat in their host, you too can turn a host into your home device! It doesn’t give you full admin access, but effectively lets you have three MARKs at all times to your home host. As long as you don’t perform illegal actions on the admin, the host spiders or any files therein, they’re likely to let you do as you please. The Foundation is still a dangerous place, even for you. So why take a host as your home?

For one, whenever you use a host as a home ‘device’ you can generate a temporary Matrix Condition Monitor of 8 boxes. This is for all intents and purposes a device without a physical location. It interacts with most Matrix mechanics as a device with a DR equal to Host Rating and a static portion of the host’s Matrix Attribute array. You may load AI-written cyberprograms and advanced programs in virtual program slots; there are (Host Rating / 2) of them for you to use, but remember that you must load into one yourself as an AI! If you take Matrix damage on this virtual device, the damage can only be repaired with the Medic advanced program or rebooting.

In addition, there are lifestyle requirements and higher Rating hosts tend to be too scrutinized to be safe places to dwell for most AIs. Do note that while living in a host has the great benefit of never being truly bricked (and thus letting you add the host’s Firewall to all your tests to resist Essence Loss), the admins are typically not happy about demiGODs banging on their door. If you ever get hit with Grid Convergence while using a host as a home, consider that lifestyle burned and lost. Doesn’t matter how much effort you put into it, chummer. Also, the host’s Matrix Attributes aren’t configurable nearly as easily; it’s a static array that you have no control over without a risky Foundation trip.

Minimum Lifestyle Host Rating Stat Array* Squatter - 500 ¥ 1 1111 Low - 2,000 ¥ 2 2111 Medium - 5,000 ¥ 3 3211 High - 10,000 ¥ 4 4321

*(The exact allocation of the Matrix Attributes is chosen when the lifestyle’s acquired.) (If anyone wants to know why this is not in use, attempts to balance effectively invulnerable hosts to very much destroyable cyberdecks proved to be a migraine; always being able to restart a persona with full stats opposed to higher but riskier physical device attributes sounded cool before we ran the numbers and concluded that 10,000 Nuyen can get you a competitive double digit Sleaze deck.)

Subcategory: Haven Mechanics
Archetype: Matrix
Topic: AI