Second, the exclusive killer feature of the public security and judicial system
Scenario 1: Intelligent review of the criminal evidence chain and early warning of contradictions
Pain point record:
The prosecutor needs to manually check the logical contradictions in 30+ pieces of evidence (such as the time conflict "19:00 at home vs 19:05 at the scene of the crime"), which is easy to lead to the withdrawal of investigation due to omissions.
Our solution:
Activate conflict detection by asking a question:
(1) Enter the → "Compare the space-time trajectory of suspect A on 2023-05-01".
Automatically output a list of time, place, and contradictions in transcription/monitoring
②
Ask the question "Checking the integrity of the chain of physical evidence" →
Returns the type of missing evidence (e.g., the tool used to commit the crime is not traced)
(3) Question the → of "Locating Witness B's Testimony and Surveillance Conflict".
Highlight contradictory paragraphs and highlight the source of the evidence
Effect:
The time taken to review the evidence was reduced from three days to two hours
Scenario 2: Multi-source data intelligent mining and AI forensic assistant
Pain point record:
The police need to manually connect clues (such as phishing emails→ dark web forums→ Bitcoin addresses) from multi-source data such as emails and chat records, and the cross-platform association takes more than 1 month.
Our solution:
Data penetration through natural language instructions:
(1) Ask the question "Search for the Telegram group associated with Bitcoin address X" →
Output the group name, chat history keyword, and associated timeline
(2) Enter the "Track the latest scam variants in phishing emails" →
Returns the change trend and propagation path of speech features
(3) Ask for "Locate the money laundering instructions corresponding to the code word 'Apple'→ Analyze the cross-platform instruction execution chain
Effect:
Lead response time from 1 month → real-time
Scenario 3: The judge intelligently retrieves the court's historical sentencing references
Pain point record:
Judges need to manually go through the file or use the basic search system to find the court's judgments in similar cases (such as "robbery causing serious injury"), which takes up to 2 hours and is difficult to accurately match the key sentencing circumstances (voluntary surrender, compensation and understanding, etc.).
Our solution:
Describe the case through natural language and accurately match historical precedents:
(1) Enter the → "Enquire about cases of robbery causing serious injury and compensation forgiveness in the past three years" The distribution of sentencing intervals and the list of typical cases will be automatically output
(2) Ask the question "Compare the difference between sentences for recidivists and non-recidivists in drug offences"→ Generate a sentence comparison table
(3) Ask about "Search for sentencing in intentional injury cases with XX circumstances (such as special defense)"→ Return to the summary of the top 5 judgments in similarity and the core sentencing reasons
Effect:
The search time for similar cases ranges from 2 hours → 10 minutes
Scenario 4: Intelligent analysis system for electronic evidence in criminal cases
Pain point record:
When a city's public security investigates a cross-border online fraud case, it needs to locate key evidence of the crime (such as instructions for the transfer of stolen money, dialogue on the division of labor between gangs) from the discovered terabyte-level chats and other communication data. Traditional keyword screening takes more than 72 hours, and it is difficult to identify encrypted code words and associated people.
Intelligent Case Handling Solution:
Penetrate massive amounts of data with natural language commands:
(1) Enter "Extract the conversation between the main culprit and the capital team about the "Apple order" from January to June 2023→ Automatically output the encrypted transaction instruction decryption text (e.g., "send 5 boxes of red Fuji" = transfer 500,000 yuan) and mark the location of the chat history
(2) Command "Link all chat accounts that use the code word "Warehouse Acceptance"→ Generate a relationship map and communication timeline of the people involved
Effect:
Key evidence mapping time ranged from 3 days → 20 minutes
Scenario 5: Intelligent analysis of massive emails in commercial bribery cases
Pain point record:
When a municipal regulatory commission investigated a bribery case of a pharmaceutical company, it faced the problem of screening massive email evidence:
Ambiguity:
Three types of key evidence need to be accurately located from 100,000+ emails;
(1) Rebate ratio (e.g. "5% rebate according to the contract amount"),
(2) bribery disguise (transfer in the name of "academic sponsorship" or "training fee"),
(3) Specify specific equipment parameter directives (e.g., "64 detectors must be procured").
Inefficiency:
Manual screening takes more than 2 weeks, and it is easy to miss cross-email association clues (for example, email A agrees on the rebate ratio, and email B performs a transfer operation).
Our solution:
Through natural language instructions, penetrating analysis is realized, and a complete evidence chain of "briber-bribe-bribe-fund flow" is automatically constructed.
(1) Enter "Analyze 'Academic Sponsorship' Emails" → Identify highly consistent templated sponsorship emails (e.g., 12 hospital leaders received invitations with similar content), but no emails to participate in academic activities were found, indicating an exception.
(2) Enter "Locate the email that requires the specified supplier and trace the decision-making chain" → Capture the supplier's email: "Customize the CT machine parameter list for you, please refer to the attachment", trace the tampering record of the associated parameters, and the bid winning notice shows the only supplier that meets the customized parameters, prompting an exception.
Effect:
The generation time of key evidence sets has been greatly shortened from 14 days for manual screening to 4 hours for intelligent analysis by the system, greatly improving the efficiency of case handling and the integrity of evidence.
