Classifying Anomalous Objects is Mostly About Not Lying to Yourself
Most people come into SCP classification thinking it is a spectrum from harmless to world-ending. It is not. The system is messier than that, and the real skill is recognizing which file is lying to you about what it actually does. Before you start slapping labels on things, you need to understand what each class actually means, because the definitions are not as clean as they look on a wiki page. Safe objects are not safe in the way you would think. The label simply means containment is straightforward once you know how to contain it. An object that turns everyone in a three-meter radius into glass is still classified Safe if you put it in a sealed lead box and write down "do not open the box." The object itself does nothing harmful when properly contained. That distinction trips up almost every first-time classifier. Euclid covers anything where the outcome is not reliably predictable. This includes objects with autonomous behavior, objects whose effects change depending on who interacts with them, or objects that occasionally defy containment without any external provocation. The vast majority of SCPs fall here, and the Foundation keeps them in Euclid because we do not have enough data to guarantee consistent results.
Keter is not synonymous with dangerous or difficult. It means the containment procedures are actively failing or require constant, resource-heavy maintenance to hold the object in place. A Keter object might be trivially easy to neutralize if we stopped caring about preserving it for study. That distinction matters more than people realize during containment reviews. Thaumiel objects exist primarily to contain other objects. They are rare, usually deeply problematic themselves, and almost always classified Keter regardless of their utility because the moment you stop maintaining them, multiple other SCPs escape. Archon is the newest class and refers to objects that should not be contained because containment would cause worse outcomes, like triggering a global anomaly or destroying evidence needed to understand a larger threat. It is essentially a "do not touch this, we are just watching it" designation. I learned this the hard way during a routine reclassification review back in 2019. We had an object labeled Safe that I had been managing for two years. It was a metallic sphere roughly forty centimeters in diameter. You closed a containment door around it and nothing happened. Boring. Then my team noticed a pattern: every time a Level 3 or above researcher entered the containment chamber, the sphere emitted a low-frequency pulse that caused localized temporal distortion within a two-meter radius. The object was not doing anything on its own. It was reacting to our presence. Changing the classification from Safe to Euclid immediately required rewriting three containment protocols and installing motion-activated dampeners. I wasted six weeks on that one because I assumed an object that sat still was simply stable. Objects that sit still are often just waiting for the wrong conditions.
The Actual Process of Classification
Classification is not a decision made in a single meeting. It is a living document that gets revised every time new data contradicts the current label. The formal process starts with a containment assessment. You need complete procedures written down for the object's baseline state. After that comes an interaction trial, which usually involves remote probes first, then d-Class subjects at a distance, then personnel in protective gear, each level adding data until you either hit a hard limit or gather enough information to make a provisional designation. The review board votes, but the vote is not final. Any researcher can request a reclassification by submitting anomalous behavior logs that contradict the current class. This happens far more often than the public-facing documents suggest. The Foundation is full of objects with mismatched labels that got stuck because nobody bothered filing the paperwork to change them. When I wrote containment reviews, I developed a simple internal check: can you describe a scenario where this object gets loose and what happens next? If your answer is "we would probably contain it quickly," that is a Safe object. If your answer is "we have no idea what it would do" or "containment would likely fail on the second day," you are looking at Euclid or Keter territory. The difference between those two comes down to predictability, not severity. A Keter object that does exactly what you expect when you follow the procedures is still Keter if the procedures break down under normal wear and tear.
Get the Full Details
Common Mistakes That Cost People Months of Work
The most frequent error is equating danger with class. A Keter label does not automatically mean an object will kill everyone. It means containment is fragile. Conversely, a Safe object can be catastrophically lethal if someone forgets to follow procedure. I have seen Senior Researchers argue for hours over whether an object should be Euclid or Keter when the real question was simply whether the current containment budget could handle a failure scenario. Those debates almost always resolve to Keter, which inflates the Keter count without meaningfully changing how we treat the object day to day. Another mistake is ignoring the containment procedure when assigning a class. The class describes the object's behavior, not our ability to handle it. A reality-bending object that occasionally spawns hostile entities is Euclid even if we have spent fifty million credits building a containment chamber specifically designed to trap those entities. The class stays the same. The budget increases. That is normal. Thaumiel classification is also wildly misused. I have seen objects labeled Thaumiel that were merely useful, not functionally critical to containing other SCPs. The threshold is high. The object must be actively preventing another SCP from causing harm, and removing it from service would create a credible containment failure. Most "Thaumiel" proposals get rejected at the first review because the proposed object is convenient, not indispensable.
What This System Actually Fails At
The classification system breaks down completely with self-modifying or cognitohazardous objects. An object that changes its own properties faster than we can observe it does not fit neatly into any class. These objects usually default to Euclid while researchers scramble to build faster observation tools, but the label is essentially a placeholder at that point. Archon classifications have solved some of these edge cases, but they are controversial because they often require admitting that the Foundation cannot control something, which creates institutional resistance. Another blind spot is multi-phase objects. I worked on a case where an SCP behaved as Safe under normal temperature conditions and as Keter when ambient temperature dropped below ten degrees Celsius. We kept it in a climate-controlled chamber and classified it Euclid because the temperature variable introduced enough unpredictability to warrant it. Some review boards wanted Keter. Some wanted Safe. The object was neither. It was just an object that required environmental monitoring as part of its containment, and the classification system gave us no clean label for that scenario. The paperwork required to change a classification is another bottleneck. Requiring formal review for every reclassification means objects sit with outdated labels for months or years while researchers wait for approval. I have seen Keter objects reclassified as Safe based on six months of clean data, then left as Keter for another year because the review board was backlogged. The label lost its meaning entirely.
A Practical Workflow for Classifying New Objects
Start by documenting baseline containment. Write down exactly what you do to keep the object sealed and what happens if you stop. If the answer is "nothing bad happens immediately," you are already leaning toward Safe. Then run interaction trials with increasing proximity until you either reach the point of no return or exhaust your trial parameters. Record everything. Changes in object behavior during trials are where reclassification requests come from. When you have enough data, draft a classification proposal and include the specific conditions that led you there. Do not just write "Euclid." Write "Euclid because interactions above two meters produce unpredictable effects lasting approximately forty seconds, and containment failure probability increases by sixty percent under stress conditions." That level of detail forces you to actually understand the object instead of just assigning a label and moving on. Expect the review board to push back. They will ask for more trials, longer observation periods, and cost estimates for alternative containment strategies. That is normal. If you cannot justify the class with data, you will be asked to collect more. There is no shortcut around that. The objects do not care how rushed you are.
Reclassification is not a failure. It is the system working as intended. An object you classified as Safe last year can become Euclid this year if someone ran a test you did not think of. The goal is not to get it right the first time. The goal is to catch when you are wrong and fix the record before someone gets hurt because they trusted a label that no longer matched reality.
