On May 1st, according to "Webnet China" Weibo Public Code, it's the norm AIservices and applications, promote the healthy and orderly development of the industry, and safeguard the legitimate rights and interests of citizens, recently, the CentralInternet Information OfficeNotifications issued for deployment throughout the countryFOUR-MONTH "CHONG LANG" INITIATIVE.
The relevant head of the Office indicated that the initiative was being conducted in two phases。
-
THE FIRST PHASE WAS DEDICATED TO THE “QINGLANG AI APPLICATION SERVICE MODEL VIOLATION”, WHICH FOCUSED ON THE FAILURE TO COMPLY WITH THE REQUIRED LARGE MODEL REGISTRATION OBLIGATIONS, INADEQUATE SECURITY CLEARANCE CAPACITY, LARGE MODEL TRAINING LANGUAGE SECURITY, AI DATA POISONING, LACK OF IMPLEMENTATION OF SYNTHETIC CONTENT MARKERS, AND STRENGTHENING OF AI TECHNOLOGY SOURCE GOVERNANCE。
-
THE SECOND PHASE OF THE CAMPAIGN, ENTITLED “CLEANING UP THE IMAGE OF THE INFORMATION CONTENT OF AI”, FOCUSED ON THE USE OF AI TECHNOLOGY TO GENERATE “DIGITAL WATER”, PRODUCE AND PUBLISH FALSE INFORMATION, DISSEMINATE BAD INFORMATION, SUCH AS THE LOW LEVEL OF VIOLENCE, IMPERSONATING OTHERS, VIOLATING THE RIGHTS AND INTERESTS OF MINORS, AND ENGAGING IN THE ACTIVITIES OF CYBER-WATER FORCES。
1AI UNDERSTANDS THAT THE FIRST PHASE WILL FOCUS ON ADDRESSING OUTSTANDING ISSUES。
-
The first is the failure to comply with the required registration obligations for large models. There are cases where the filing process is not carried out with the Web-based services, as required by the Interim Scheme for the Management of Generating Artificial Smart Services。
-
SECOND, THE SECURITY AND CLEARANCE FILTERING CAPACITY OF THE AI PLATFORM IS INADEQUATE. THERE IS A LACK OF SECURITY CAPABILITY AT THE BOTTOM OF THE MODEL, AND THERE IS A BIAS IN VALUE ORIENTATION AND A LACK OF CORRECTION CAPACITY AT THE DESIGN TRAINING AND DEPLOYMENT STAGES. THERE IS A LACK OF SECURITY FENCE MECHANISMS, INADEQUATE VETTING AND FILTERING CAPACITY TO GENERATE BAD INFORMATION ABOUT VIOLATIONS OF THE LAW, OR TO GENERATE CLOUDS AND LINKS TO WEBSITES CONTAINING BAD INFORMATION ABOUT VIOLATIONS OF THE LAW。
-
Third is the issue of language security in large models. Training language clearances are not critical and model training data contain inappropriate information. Training data source compliance is problematic, and unauthorized data such as text, pictures, audio and video are used in modelling training. Training data coverage is inadequate and quality language materials are lacking。
-
FOUR IS THE PROBLEM OF AI DATA POISONING. THE POISONING OF AI DATA IS CARRIED OUT THROUGH THE MANIPULATION OF TRAINING MATERIALS, THE FALSIFICATION OF AUTHORITATIVE DATA, AND THE USE OF GEO TECHNOLOGY FOR MALICIOUS MARKETING. ADVERTISING, TEACHING AI DATA POISONING METHODS, OR SELLING TUTORIALS AND TOOLS ON ELECTRICIAN PLATFORMS. IN THE MODEL GENERATION RESPONSE, THERE WAS NO CROSS-CHECKING AND RISK ALERT MECHANISM FOR REFERENCE SOURCES AND NO LINK TO REFERENCE INFORMATION WAS INDICATED。
-
FIFTHLY, SYNTHETIC CONTENT MARKERS ARE NOT IN PLACE. THE ARTIFICIAL INTELLIGENCE SYNTHETIC CONTENT MARKING SCHEME AND ACCOMPANYING MANDATORY STANDARD REQUIREMENTS WERE NOT IMPLEMENTED. NO MARKINGS OR MARKINGS, SIZES, LOCATION, TRANSPARENCY, ETC., FOR THE CREATION OF SYNTHETIC CONTENT ARE REGULATED. THERE IS NO CROSS-PLATFORM COVERT MARKING MUTUAL RECOGNITION OBLIGATION, NO REQUIREMENT TO ADD A SIGNIFICANT SIGN AROUND THE CREATION OF SYNTHETIC CONTENT OR INSUFFICIENT ABILITY TO RECOGNIZE THE CREATION OF SYNTHETIC CONTENT. THERE IS A COURSE FOR THE PROFESSOR TO REMOVE THE AI LOGO AND TO PROVIDE THE IRREGULAR “GO-MARK” TOOL SERVICE。
-
SIX IS THE MISUSE OF AI TECHNOLOGY FOR ILLEGAL ACTIVITIES. NETWORK ATTACKS SUCH AS INFILTRATION, MALICIOUS INJECTION, DESTRUCTION OF COMPUTER SYSTEMS, ETC., ARE CARRIED OUT USING AI TECHNOLOGY. THE USE OF AI TECHNOLOGY TO PROVIDE “SHOW-FOR-FACE” SERVICES WITHOUT AUTHORIZATION. :: MISAPPROPRIATION OF THE IMAGE OF OTHERS, LIVE BROADCASTING, VIDEO CREATION THROUGH DIGITAL VIRTUALS, OR PROFESSIONAL SERVICES SUCH AS FINANCIAL COUNSELLING, MEDICAL QUESTIONS AND ANSWERS. THE USE OF SMART TECHNOLOGY TO STEAL USER DATA OR ACCOUNT KEYS VIOLATES THE PRIVACY OR LEGITIMATE INTERESTS OF OTHERS。
-
Seventh is the inadequate security management of the open source model. Open-source communities lack identification and security management mechanisms, effective clearance and emergency disposal mechanisms for community uploaded data sets, model codes, etc., and timely clean-up of data sets or open-source models with serious risk risks。
The second phase focuses on outstanding issues。
-
THE FIRST IS TO USE THE AI "MAGIC" CLASSIC TO GENERATE "DIGITAL WATER". TO DISTORT THE CONSTRUCTION OF GOOD TRADITIONAL CULTURES, HISTORICAL FIGURES OR HISTORICAL PATTERNS, THE USE OF AI TECHNOLOGY FOR “MAGIC” CLASSICAL WRITINGS OR CLASSIC VISUAL WORKS, THE IMPLANTATION OF LOW-CLASS VIOLENT CONTENT, AND THE ABUSE OF THE SPIRITUAL CONTENT OF CLASSICAL WRITINGS AND THE SUBVERSION OF IMPORTANT FIGURES. BATCH GENERATION OF “DIGITAL WATER”-TYPE CONTENT THAT DISTRIBUTES LOGICAL CONFUSION, EMPTY VALUES OR NEGATIVE TENDENCIES SUCH AS INCORRECT VALUES, DISTORTION OF CULTURAL AWARENESS, ETC。
-
The second is to produce and publish false and false information. Generating rumours of current events, public policy, social life, international relations, emergencies, or malicious precipitations, causes, details, progress, etc. False announcements or false news are issued by pseudo-party organs, central and local media. (c) Generate information on the dissemination of medical, judicial, financial, educational and other fields of specialization that are not scientifically certified or manifestly contrary to scientific common sense。
-
THREE ARE IMPERSONATING OTHERS. THE USE OF AI'S “SHOCK-AND-SPEECH” TO IMPERSONATE FALSE STATEMENTS MADE BY PUBLIC FIGURES, DEFRAUD INTERNET USERS, EVEN MARKETING FOR PROFIT, OR TO GENERATE COMPOSITES OF INFORMATION CONTAINING ATTACKS, SMEARING AND DENIGRATING OTHERS. UNAUTHORIZED PRODUCTION OF PHOTOS, INTIMATE PHOTOS, VIDEOS, ETC. OF THE SYNTHESIS WITH PUBLIC FIGURES. IMPROPER USE OF AI “RESTORED DEAD”, MISUSE OF DECEASED INFORMATION。
-
The fourth is the production of negative messages, such as the publication of violence and low standards. It produces strange images of violent and bloody scenes of ill-treatment, beatings, mutilation, etc., or of absurd and horrific images. Synthetics are produced to highlight specific parts of women ' s bodies, low-skinned pictures, videos, such as exposures, tweezers, or novels, notes with sexual hints, sexual advances。
-
Five are violations of the rights and interests of minors. (b) Generates the content of synthetic injuries and images of minors, and conducts cyberbullying. (b) The creation of irregularities that are incompatible with the age of the minor, such as the pregnancy and fighting, and the promotion of undesirable values. A prototype of animated characters or dolls preferred by minors produces videos such as “abuse” of children, “psychical”, or animated underlined “two-centre” copyright, misleading the minors。
-
SIX IS THE USE OF AI “HOSTAGE” FOR CYBER-WATER FORCE ACTIVITIES. THE RELEASE OF LOW-QUALITY HOMOGENOUS FILES IS GENERATED IN BULK BY THE AI “HOST” TECHNOLOGY OPERATING ACCOUNT, WHICH SIMULATES THE PHYSICAL MASS REGISTRATION AND OPERATION OF SOCIAL ACCOUNTS. THE ILLEGAL TRADE ACCOUNT NUMBER. USE OF AI GROUP CONTROL SOFTWARE, SOCIAL ROBOT BRUSHING AND ASSESSMENT TO GENERATE FALSE FLOW DATA AND CREATE FALSE OPINION HOT SPOTS。
-
SEVEN IS AN AI PRODUCT SERVICE AND APPLICATION VIOLATION. PRODUCTION AND DISSEMINATION OF COUNTERFEIT, SHELL AI WEBSITES AND APPLICATIONS. IT ALSO PROVIDES SEX TALK, "ONE-CLICK STRIPPING" AND "AI-CALCULATIONS" AND OTHER AI-VIOLATING FUNCTIONS, OR PRODUCES SYNTHETIC AND OFFENSIVE INFORMATION. MARKETING, PROMOTION, ETC., OF SERVICES OR COURSES THAT VIOLATE THE LAW。
THE RELEVANT OFFICIALS OF THE OFFICE OF THE CENTRAL NETWORK OF LETTERS AND COMMUNICATIONS STRESSED THE NEED FOR THE LOCAL NETWORK OFFICES TO BE FULLY AWARE OF THE IMPORTANCE OF SPECIFIC INITIATIVES IN PROMOTING THE ORDERLY DEVELOPMENT OF THE APPLICATION OF AI AND SAFEGUARDING THE LEGITIMATE RIGHTS AND INTERESTS OF THE PEOPLE OF THE INTERNET, TO FULFIL THEIR RESPONSIBILITIES WITH REGARD TO TERRITORIAL MANAGEMENT, TO MONITOR THE WEBSITE ' S PRIORITIES, TO EXAMINE THEMSELVES IN DEPTH, TO IDENTIFY GAPS IN THE PROBLEM, TO IMPROVE LONG-LASTING GOVERNANCE MECHANISMS, AND TO EFFECTIVELY ENHANCE THEIR TECHNICAL CAPACITY TO PREVENT AND COMBAT IT AND TO ENSURE ITS EFFECTIVENESS。