본문 바로가기 주메뉴 바로가기
국회도서관 홈으로 정보검색 소장정보 검색

결과 내 검색

동의어 포함

목차보기

Title page 1

Contents 3

1. Executive summary 4

1.1. Developments in generative AI 4

1.2. Growth in agentic AI and multi-agent systems 4

1.3. Investments, partnerships and competition for talent 5

1.4. Consumer issues 6

1.5. Continued monitoring is needed 7

2. The continued rise of Gen AI 8

2.1. Generative AI models in 2025 8

2.1.1. Benchmarks indicate significant improvements in new models 10

2.1.2. New types of models are being released 11

2.2. AI applications in 2025 14

2.2.1. AI increasingly draws on personalised data 14

2.2.2. AI is being integrated across digital ecosystems 14

2.2.3. AI could change how the internet is accessed 15

2.2.4. Firms are exploring monetisation strategies for AI apps 17

2.2.5. Firms have released new AI image and video apps 17

3. Agentic AI since March 2025 18

3.1. What are AI agents? 18

3.2. Agentic product announcements and releases 21

3.3. Agentic frameworks 23

3.4. Continued monitoring is required 24

3.5. International monitoring 26

4. Investments, acquisitions, and partnerships 27

4.1. Significant investments continue to be made in the AI supply chain 28

4.1.1. Significant investments are being made in AI infrastructure 28

4.1.2. Key players are investing to vertically integrate and self-supply compute 31

4.2. Strategic partnerships, mergers and acquisitions 34

4.2.1. Partnerships for investment in AI infrastructure 34

4.2.2. Interdependencies across the AI supply chain 36

4.2.3. Competition for technical experts 37

5. Consumer risks related to AI 40

5.1. Use of consumer data 41

5.2. Potential misleading or deceptive conduct related to AI 43

5.2.1. Use of generative AI to facilitate false representations 44

5.2.2. Use of AI chatbots in customer service 44

5.2.3. 'AI-washing' practices 45

5.3. Risk of AI-generated fake and manipulated reviews 46

5.4. Use of AI in manipulative design practices 48

5.4.1. Hypernudging 48

5.5. Use of AI in scams 49

5.6. Protecting consumers from AI harms 52

Figures 5

Figure 1.1. Capital expenditure by select digital platforms, 2020-2025 5

Figure 1.2. Interdependencies in the AI supply chain 6

Figure 2.1. A selection of notable AI model releases in 2025 10

Figure 2.2/Figure 2.3. Example of a simulated environment using Google's Genie 3 13

Figure 3.1. An augmented LLM AI agent 19

Figure 3.2. Multi-agent systems can be set up in various architectures 20

Figure 3.3. Examples of agentic framework releases by major AI firms since March 2025 24

Figure 4.1. Comparison of digital platforms' capital expenditure and Australian spending 29

Figure 4.2. Capital expenditure by select digital platforms, 2020-2025 29

Figure 4.3. Vertical integration across the AI stack 32

Figure 4.4. Interdependencies in the AI supply chain 37

Figure 5.1. Consumer views regarding consent to use personal data to train AI 41

Figure 5.2. Screenshots of landing page 50

Figure 5.3. Screenshots of landing page 51

Figure 5.4. Screenshot from alternate landing page 51

Boxes 9

Box 2.1. Examples of new AI model releases 9

Box 2.2. Benchmarks provide an indication of model quality 11

Box 3.1. Case study - ChatGPT Instant Checkout uses OpenAI's open-source Agentic Commerce Protocol 23

Box 4.1. AI infrastructure in Australia 33

Box 4.2. Competition authorities internationally continue to scrutinise mergers, acquisitions and partnerships 39

Box 5.1. ACCC action on alleged misleading Microsoft 365 subscription 43

Box 5.2. US authorities are scrutinising claims about advanced AI functionality 45

Box 5.3. Selected international legislative and regulatory developments on AI-generated fake reviews 47

Box 5.4. General prohibition on unfair trading practices to be introduced in Australia 48

Box 5.5. Example AI-washing investment scam website 50

Box 5.6. Selected recent AI regulatory and policy developments in Australia 53