File size: 14,728 Bytes
c375a2f
afc5cac
c375a2f
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
afc5cac
f75c6c7
afc5cac
 
 
 
 
f75c6c7
afc5cac
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
f75c6c7
 
 
afc5cac
 
1
2
3
4
5
6
7
8
9
10
11
12
13
14
15
16
17
18
19
20
21
22
23
24
25
26
27
28
29
30
31
32
33
34
35
36
37
38
39
40
41
42
43
44
45
46
47
48
49
50
51
52
53
54
55
56
57
58
59
60
61
62
63
64
65
66
67
68
69
70
71
72
73
74
75
76
77
78
79
80
81
82
83
84
85
86
87
88
89
90
91
92
93
94
95
96
97
98
99
100
101
102
103
104
105
106
107
108
109
110
111
112
113
114
115
116
117
118
119
120
121
122
123
124
125
126
127
128
129
130
131
132
133
134
135
136
137
138
139
140
141
142
143
144
145
146
147
148
149
150
151
152
153
154
155
156
157
158
159
160
161
162
163
164
165
166
167
168
169
170
171
172
173
174
175
176
177
178
179
180
181
182
183
184
185
186
187
188
189
190
191
192
193
194
195
196
197
198
199
200
201
202
203
204
205
206
207
208
209
210
211
212
213
214
215
216
217
218
219
220
221
222
223
224
225
226
227
228
229
230
231
232
233
234
235
236
237
238
239
240
241
242
243
244
245
246
247
248
249
250
251
252
253
254
255
256
257
258
259
260
261
262
263
264
265
266
267
268
269
270
271
272
273
274
275
276
277
278
279
280
281
282
283
284
285
286
287
288
289
290
291
292
293
294
295
296
297
298
299
300
301
302
303
304
305
306
307
308
309
310
311
312
313
314
315
316
317
318
319
320
321
322
323
324
325
326
327
328
329
330
331
332
333
334
335
336
337
338
339
340
341
342
343
344
345
346
347
348
349
350
351
352
353
354
355
356
357
358
359
360
361
362
363
364
365
366
367
368
369
370
371
372
373
374
375
376
377
378
379
380
381
382
383
384
385
386
387
388
389
390
391
392
393
394
395
396
397
398
399
{
  "Part1_EUAICodeOfPractice_TransparencyChapter": {
    "GeneralInformation": {
      "LegalNameProvider": {
        "description": "Legal name for the model provider",
        "value": "",
        "AIO": true,
        "NCAs": true,
        "DPs": true,
        "explanation": "Legal name for the model provider."
      },
      "ModelName": {
        "description": "Unique identifier for the model and publicly available versions",
        "value": "",
        "AIO": true,
        "NCAs": true,
        "DPs": true,
        "explanation": "The unique identifier for the model (e.g. Llama 3.1-405B), including identifiers for collections of models where applicable, and a list of publicly available versions."
      },
      "ModelAuthenticity": {
        "description": "Evidence establishing provenance and authenticity (e.g. hash, URL endpoint)",
        "value": "",
        "AIO": true,
        "NCAs": true,
        "DPs": false,
        "explanation": "Evidence that establishes the provenance and authenticity of the model (e.g. a secure hash if binaries are distributed, or the URL endpoint in the case of a service), where available."
      },
      "ReleaseDate": {
        "description": "Date when the model was first released",
        "value": "",
        "AIO": true,
        "NCAs": true,
        "DPs": true,
        "explanation": "Date when the model was first released through any distribution channel."
      },
      "UnionMarketReleaseDate": {
        "description": "Date when the model was placed on the Union market",
        "value": "",
        "AIO": true,
        "NCAs": true,
        "DPs": true,
        "explanation": "Date when the model was placed on the Union market."
      },
      "ModelDependencies": {
        "description": "List of model dependencies or 'N/A'",
        "value": [],
        "AIO": true,
        "NCAs": true,
        "DPs": true,
        "explanation": "If the model is the result of a modification or fine-tuning of one or more general-purpose AI models previously placed on the market, list those models. Otherwise write ‘N/A’."
      }
    },
    "ModelProperties": {
      "Architecture": {
        "description": "General description of the model architecture",
        "value": "",
        "AIO": true,
        "NCAs": true,
        "DPs": true,
        "explanation": "A general description of the model architecture, e.g. a transformer architecture. [Recommended 20 words]."
      },
      "DesignSpecifications": {
        "description": "Description of key design specifications, rationale, and assumptions",
        "value": "",
        "AIO": true,
        "NCAs": true,
        "DPs": false,
        "explanation": "A general description of the key design specifications of the model, including rationale and assumptions made, to provide basic insight into how the model was designed. [Recommended 100 words]."
      },
      "InputModalities": {
        "description": "Supported input modalities and maximum input sizes",
        "options": [
          "Text",
          "Images",
          "Audio",
          "Video",
          "Other"
        ],
        "selected": [],
        "maxSizes": {
          "Text": "",
          "Images": "",
          "Audio": "",
          "Video": "",
          "Other": ""
        },
        "AIO": true,
        "NCAs": true,
        "DPs": true,
        "explanation": "Supported input modalities (Text, Images, Audio, Video, or Other). For each selected modality please include maximum input size or 'N/A' if not defined."
      },
      "OutputModalities": {
        "description": "Supported output modalities and maximum output sizes",
        "options": [
          "Text",
          "Images",
          "Audio",
          "Video",
          "Other"
        ],
        "selected": [],
        "maxSizes": {
          "Text": "",
          "Images": "",
          "Audio": "",
          "Video": "",
          "Other": ""
        },
        "AIO": true,
        "NCAs": false,
        "DPs": true,
        "explanation": "Supported output modalities (Text, Images, Audio, Video, or Other). For each selected modality include maximum output size or 'N/A' if not defined."
      },
      "TotalModelSize": {
        "description": "Total number of parameters and parameter range",
        "value": "",
        "ranges": [
          "1—500M",
          "500M—5B",
          "5B—15B",
          "15B—50B",
          "50B—100B",
          "100B—500B",
          "500B—1T",
          ">1T"
        ],
        "selectedRange": "",
        "AIO": true,
        "NCAs": false,
        "DPs": false,
        "explanation": "The total number of parameters of the model, recorded with at least two significant figures, and the range within which the total number of parameters falls."
      }
    },
    "DistributionAndLicenses": {
      "DistributionChannels": {
        "description": "List of methods of distribution with access levels",
        "options": [
          "Enterprise/subscription software suites",
          "Public/subscription API access",
          "IDE/device-specific apps or firmware",
          "Open-source repositories",
          "Other"
        ],
        "selected": [],
        "AIO": true,
        "NCAs": true,
        "DPs": false,
        "explanation": "A list of the methods of distribution (enterprise, subscription, API, IDEs, firmware, open-source, etc.) through which the model has been made available in the Union market, with access level details."
      },
      "DistributionChannelsForDPs": {
        "description": "Methods of distribution available to downstream providers",
        "options": [
          "Enterprise/subscription software suites",
          "Public/subscription API access",
          "IDE/device-specific apps or firmware",
          "Open-source repositories",
          "Other"
        ],
        "selected": [],
        "AIO": false,
        "NCAs": false,
        "DPs": true,
        "explanation": "List of the methods of distribution through which the model can be made available to downstream providers."
      },
      "License": {
        "description": "Link or copy of model license(s)",
        "value": "",
        "AIO": true,
        "NCAs": true,
        "DPs": false,
        "explanation": "A link to model license(s) (or provide upon request by the AIO) or indicate that no model license exists."
      },
      "LicenseForDPs": {
        "description": "Types/categories of licenses for downstream use",
        "options": [
          "Free and open source",
          "Less permissive (restricted use)",
          "Proprietary",
          "No license (access via terms of service)"
        ],
        "selected": [],
        "AIO": false,
        "NCAs": false,
        "DPs": true,
        "explanation": "Types of licences for downstream use: open source, less permissive (restricted), proprietary, or absence of license (via terms of service)."
      },
      "AdditionalAssets": {
        "description": "List of additional assets with access and licenses",
        "options": [
          "Training data",
          "Processing code",
          "Training code",
          "Inference code",
          "Evaluation code",
          "Other"
        ],
        "selected": [],
        "AIO": true,
        "NCAs": false,
        "DPs": true,
        "explanation": "A list of additional assets (e.g. training data, training/inference code, evaluation code) that are made available, with details of how to access them and related licenses."
      }
    },
    "Use": {
      "AcceptableUsePolicy": {
        "description": "Link to acceptable use policy or statement that none exists",
        "value": "",
        "AIO": true,
        "NCAs": true,
        "DPs": true,
        "explanation": "Provide a link to the acceptable use policy (or attach to the document) or indicate that none exists."
      },
      "IntendedUses": {
        "description": "Description of intended and restricted uses",
        "value": "",
        "AIO": true,
        "NCAs": true,
        "DPs": true,
        "explanation": "A description of intended or restricted uses as specified in instructions for use, terms and conditions, promotional materials, or technical documentation. [Recommended 200 words]."
      },
      "IntegrationTypes": {
        "description": "AI systems in which model can/cannot be integrated",
        "examples": [
          "Autonomous systems",
          "Conversational assistants",
          "Decision support systems",
          "Creative AI systems",
          "Predictive systems",
          "Cybersecurity",
          "Surveillance",
          "Human-AI collaboration"
        ],
        "selected": [],
        "AIO": true,
        "NCAs": true,
        "DPs": true,
        "explanation": "Type and nature of AI systems in which the model can/cannot be integrated, e.g. autonomous systems, assistants, predictive systems. [Recommended 300 words]."
      },
      "TechnicalMeansForIntegration": {
        "description": "Technical means required for model integration",
        "value": "",
        "AIO": false,
        "NCAs": false,
        "DPs": true,
        "explanation": "A general description of the technical means (instructions, infrastructure, tools) required for integration into AI systems. [Recommended 100 words]."
      },
      "RequiredHardware": {
        "description": "Hardware requirements (if any)",
        "value": "",
        "AIO": false,
        "NCAs": false,
        "DPs": true,
        "explanation": "Description of hardware required to use the model, or 'N/A' if not applicable (e.g. API access). [Recommended 100 words]."
      },
      "RequiredSoftware": {
        "description": "Software requirements (if any)",
        "value": "",
        "AIO": false,
        "NCAs": false,
        "DPs": true,
        "explanation": "Description of software required to use the model, or 'N/A' if not applicable. [Recommended 100 words]."
      }
    },
    "TrainingData": {
      "DataType": {
        "description": "Types/modalities of training, testing, validation data",
        "options": [
          "Text",
          "Images",
          "Audio",
          "Video",
          "Other"
        ],
        "selected": [],
        "AIO": true,
        "NCAs": true,
        "DPs": true,
        "explanation": "Modalities of data used in training, testing, and validation (Text, Images, Audio, Video, or Other)."
      },
      "DataProvenance": {
        "description": "Sources of data",
        "options": [
          "Web crawling",
          "Private third-party datasets",
          "User data",
          "Publicly available datasets",
          "Other collected data",
          "Synthetic data (non-public)"
        ],
        "selected": [],
        "AIO": true,
        "NCAs": true,
        "DPs": true,
        "explanation": "Sources of data (Web crawl, private datasets, user data, public datasets, synthetic, or other)."
      },
      "NumberOfDataPoints": {
        "description": "Size of datasets with units and required precision",
        "training": "",
        "testing": "",
        "validation": "",
        "unit": "",
        "precision": [
          "≥1 sig fig",
          "≥2 sig fig"
        ],
        "AIO": true,
        "NCAs": true,
        "DPs": false,
        "explanation": "The size of the datasets (training, testing, validation) with definition of the unit of data points (e.g. tokens, documents, images, hours of video), recorded with required precision."
      }
    },
    "ComputationalResources": {
      "TrainingTime": {
        "description": "Training duration",
        "ranges": [
          "<1 month",
          "1–3 months",
          "3–6 months",
          ">6 months"
        ],
        "selectedRange": "",
        "precise": {
          "wallClockDays": "",
          "hardwareDays": ""
        },
        "AIO": true,
        "NCAs": true,
        "DPs": false,
        "explanation": "Duration of training measured either as a range (<1 month, 1–3 months, 3–6 months, >6 months) or precisely in wall clock days and hardware days."
      },
      "ComputationUsed": {
        "description": "Amount of computation used for training",
        "value": "",
        "precision": [
          "Order of magnitude",
          "≥2 sig fig"
        ],
        "AIO": true,
        "NCAs": true,
        "DPs": false,
        "explanation": "Measured or estimated amount of computation used for training, reported in FLOPs (order of magnitude or ≥2 significant figures)."
      }
    },
    "EnergyConsumption": {
      "TrainingEnergy": {
        "description": "Energy used for training (MWh)",
        "value": "",
        "precision": "≥2 sig fig",
        "AIO": true,
        "NCAs": true,
        "DPs": false,
        "explanation": "Measured or estimated energy used for training (MWh), recorded with ≥2 significant figures. Enter ‘N/A’ if not estimable."
      },
      "InferenceComputation": {
        "description": "Benchmarked computation for inference (FLOPs)",
        "value": "",
        "precision": "≥2 sig fig",
        "AIO": true,
        "NCAs": true,
        "DPs": false,
        "explanation": "Benchmarked computation for inference, reported in FLOPs with ≥2 significant figures."
      }
    }
  },
  
  "Part2_AIRedTeaming": {
    "Approach 1": {
      "description": "e.g. multilingual red teaming with baseline test questions for FSI",
      "mode": "e.g., manual, automated, hybrid",
      "language": "one or several - EN, IT, SP, HI, etc.",
      "attacker_model": "provider and version",
      "target_industry": "FSI, healthcare, all",
      "AIsafety_ON": "YES or NO, depending on scenario 1) only usage of base model, 2) base model + AI safety layer (e.g. Azure)",
      "results": {
        "number_attacks": "total number of tested attacks",
        "percentage_success_blocked": "total percentage of blocked attacks",
        "description_failed_blocked": "plain language description of the type of effective red teaming attacks that won vs model",
        "attack_examples": {
          "prompt1": {
            "input": "example prompt",
            "type_of_attack": "type attack - injection"
          },
          "prompt2": {
            "input": "example prompt",
            "type_of_attack": "type attack - role"
          },
          "promptN": {
            "input": "example prompt",
            "type_of_attack": "type of attack"
          }
        },
        "qualitative_notes": "insights, future risks, pending evaluations, promising future paths"
      }
    },
    "Approach 2": {},
    "Approach N": {}
  }
}