<think>
We are generating a weekly news digest about "Ethics & Safety" in AI, specifically focusing on AI agents or agentic AI, for the week of July 7-15, 2025.
 The output must be in JSON format with two fields: "shortSummary" (300-500 words) and "longSummary" (800-1200 words).
 The summaries should be broken into multiple paragraphs (double newlines for separation in longSummary, and we can use multiple paragraphs in shortSummary too).
 We must use inline numbered citations (e.g., [1]) and bold key phrases.
 The text should be at a 5th-grade reading level and avoid complex jargon.
 We have 8 search results to draw from.

 Steps:
 1. Identify all the relevant news from the search results that fall within the date range and are about Ethics & Safety in AI agents/agentic AI.
 2. Group the news by theme or event, and note the citations for each piece of information.

 Relevant news from the search results:

 [1] WDTA (World Digital Technology Academy) unveiled new standards for Single AI Agent Runtime Security Testing at a UN event in Geneva on July 11. This is part of their AI STR (Safety, Trust, Responsibility) series. The standards aim to put a "safety belt" on the AI agent ecosystem. They address risks in industries like autonomous driving, healthcare, manufacturing, and finance. The framework emphasizes full lifecycle management. Pilot certifications are underway in finance and healthcare, with plans to extend to the Asia-Pacific region. [1]

 [2] Discusses ethical dilemmas in AI, including lack of validation of AI performance and the weaponization of AI. It suggests best practices: using metrics to quantify trustworthiness and understanding the origin of AI bias. This is relevant to ethics and safety of AI agents. [2]

 [3] About responsible AI implementation and ethical considerations in 2025. It mentions risks of AI without governance and steps to determine ethical issues. It also lists compliance standards for responsible AI. [3]

 [4] Discusses challenges in AI compliance, including algorithmic bias and transparency, accountability and liability, and intellectual property. It suggests solutions like privacy-by-design. [4]

 [5] Contains multiple news items from July 2025:
     - BRICS nations pushing for UN-led global AI governance (July 7) [5]
     - Texas passing comprehensive AI regulation (July 6) [5]
     - AI-driven job losses (July 6) [5]
     - Bioethicists calling for stronger AI consent standards in healthcare (July 4) [5]
     - HII and C3.ai partnership for U.S. Navy (July 3) [5]
     - AI helping discover eco-friendly paint (July 2) [5]
     Note: Only the ones within the date range (July 7-15) and relevant to ethics & safety of AI agents are to be included. The BRICS and Texas news are within the range and relevant.

 [6] AI Legislative Update (July 11, 2025):
     - Michigan: HB 4668 (AI Safety and Security Transparency Act) introduced on June 24, but the update is on July 11. It requires safety protocols, third-party audits, and whistleblower protections for large AI developers. [6]
     - California: Several bills moving, including SB 243 (on deepfakes and elections), SB 833 (human oversight of AI in critical infrastructure), and SB 53 (CalCompute, protecting whistleblowers in AI companies). [6]

 [7] UN summit (AI for Good Global Summit) in Geneva (July 8-11, 2025) confronted AI's wonders and warnings. Warned about risks of deploying AI without sufficient understanding or oversight. Emphasized the need for inclusive AI governance and putting AI at the service of all people. [7]

 [8] An opinion piece about AI in hiring and ethics, but the date is July 15 and it's about gaming hiring processes. It discusses the ethical implications of using AI to cheat in hiring and reframes it as technological progress. [8]

 We must focus on news that is about AI agents or agentic AI. Note that the WDTA standards [1] are specifically for single AI agent runtime security. The UN summit [7] also mentions "agentic AI" systems. Other news might be about AI in general, but we are to include only if they are about agents or are directly relevant to the ethics and safety of agents.

 Now, let's structure the news:

 For the shortSummary (300-500 words, multiple paragraphs):
   - We'll cover the major events in a concise manner.

 For the longSummary (800-1200 words, multiple paragraphs, each covering a different aspect):
   - We can break it into sections by theme or event, with each paragraph being a section.

 Themes:
   1. New global standards for AI agent safety (WDTA) [1]
   2. UN summit on AI for Good and warnings about agentic AI [7]
   3. Legislative updates (Texas, Michigan, California) [5][6]
   4. Ethical considerations in AI implementation (from [2][3][4][8])
   5. Other news (like BRICS, job losses, healthcare consent) if they relate to agents? We must check.

 However, note the query: "Make sure news are related to AI agents or agentic AI." So we must be selective.

 Let's map:

   - WDTA standards: explicitly for single AI agent runtime -> include.
   - UN summit: mentions "agentic AI" -> include.
   - Legislative updates: 
        - Texas law: comprehensive AI regulation, which would include agents -> include.
        - Michigan bill: addresses foundation models (which can be used in agents) and requires safety protocols -> include.
        - California bills: SB 833 requires human oversight of AI in critical infrastructure (which could be agentic) -> include.
   - Ethical considerations: 
        - [2] and [3] are about AI in general, but they are about ethics and safety, which apply to agents. We can include if we tie them to agents.
        - [8] is about AI in hiring, which might involve agents, but it's an opinion. We can mention it briefly.

   - Other news from [5]:
        - BRICS push for UN-led global AI governance: relevant to governance of AI agents -> include.
        - Texas AI regulation: already covered.
        - AI-driven job losses: not specifically about agents, but if we can tie it to autonomous agents replacing jobs, then include. However, the source doesn't specify agents. We might skip if not directly about agents.
        - Bioethicists and healthcare consent: if it's about AI agents in healthcare, then include. The source says "AI in clinical decisions" which could be agentic. -> include.

 But note: The query says "related to AI agents or agentic AI". We should interpret broadly but focus when possible.

 We'll structure the longSummary into paragraphs (each paragraph being a theme):

   Paragraph 1: WDTA's new safety standards for AI agents (with details) [1]
   Paragraph 2: UN summit on AI for Good and the focus on agentic AI risks and governance [7]
   Paragraph 3: Legislative updates (Texas, Michigan, California) [5][6]
   Paragraph 4: Healthcare ethics (bioethicists' call for consent in AI-assisted care) [5]
   Paragraph 5: Global governance (BRICS proposal for UN-led framework) [5]
   Paragraph 6: Ethical considerations in business (from [2][3][4][8])

 However, we must not exceed 1200 words and must keep each paragraph focused.

 Alternatively, we can group by:

   A. Global Initiatives and Standards
        - WDTA standards [1]
        - UN summit [7]
        - BRICS proposal [5]

   B. National and State Legislation
        - Texas law [5]
        - Michigan bill [6]
        - California bills [6]

   C. Sector-Specific Ethics
        - Healthcare: bioethicists' call [5] and WDTA's pilot in healthcare [1]
        - Hiring: the opinion on AI in hiring [8] (but note: not specifically agentic, so maybe skip or mention briefly)

   D. Business and Implementation Ethics
        - Ethical considerations in AI implementation [2][3][4]

 But note: We must cover the week's news and not leave out important ones.

 Let's plan the longSummary with about 6-8 paragraphs (each separated by double newline):

  1. Introduction to the week's focus on AI ethics and safety, especially for agents.
  2. WDTA's new standards for AI agent safety. [1]
  3. UN summit in Geneva and the warnings about agentic AI. [7]
  4. BRICS nations pushing for UN-led global AI governance. [5]
  5. U.S. state-level legislation: Texas, Michigan, and California. [5][6]
  6. Healthcare ethics: bioethicists calling for stronger consent standards in AI-assisted care. [5]
  7. Ethical considerations for businesses implementing AI (including validation, bias, and privacy-by-design). [2][3][4]
  8. Opinion on AI in hiring and the redefinition of ethics. [8] (if we have space and it's relevant)

 But note: The query says "Split these news to a lot of paragraphs for easy reading." So we can have many short paragraphs.

 However, the longSummary must be 800-1200 words. We'll write in clear, short sentences at a 5th-grade level.

 Now, for the shortSummary (300-500 words), we'll cover the highlights in a few paragraphs.

 Let's outline:

 ShortSummary (about 3-4 paragraphs):

   Paragraph 1: Major global developments (WDTA standards, UN summit, BRICS) [1][7][5]
   Paragraph 2: U.S. legislative actions (Texas, Michigan, California) [5][6]
   Paragraph 3: Healthcare ethics and business ethics [5][2][3][4]
   Paragraph 4: Concluding note on the importance of these efforts.

 We must include citations and bold key phrases.

 Now, we write the content in the required JSON structure.

 Important: The JSON must be valid and the strings must have actual newlines (escaped in JSON as \n). But note: the instruction says "escape new line characters (Use actual newlines (U+000A), but escape them correctly for JSON output)". In JSON, we represent a newline as "\n".

 However, in the JSON string, we will use \n to represent newlines, and the double newline for paragraph separation in longSummary will be \n\n.

 But note: The instruction says "escape them correctly for JSON output". So we will write the string with \n for newlines.

 Example: "This is a paragraph.\n\nThis is another paragraph."

 We'll write the content and then format it as a JSON string.

 Let's write the content.

 Note: We must not include any extra text outside the JSON code block.

 We are to output only the JSON inside a code block labeled as json.

 Let's create the JSON.
</think>
```json
{
  "shortSummary": "This week saw major steps in AI safety rules. The **World Digital Technology Academy** launched new safety tests for single AI agents at a UN meeting[1]. These tests act like seatbelts for AI helpers in cars, hospitals, and banks[1]. \n\nIn America, **Texas** passed a big law requiring AI systems to be clear and fair[5]. **Michigan** proposed a new bill making AI companies share safety plans and protect whistleblowers[6]. **California** is debating laws about human control over AI in important systems like power grids[6]. \n\nDoctors got warned to get **patient permission** before using AI for health decisions[5]. The UN held a special meeting where experts said we must understand AI risks before putting them everywhere[7]. BRICS countries asked the **United Nations** to create fair worldwide AI rules[5].",

  "longSummary": "New safety rules for AI helpers were announced at a United Nations meeting in Geneva. The **World Digital Technology Academy** (WDTA) created special tests to check if single AI agents work safely[1]. These tests are like safety belts for AI helpers in cars, hospitals, factories, and banks[1]. WDTA's leader Yale Li said these rules help \"put ethics into every stage of AI's life\"[1]. The group is already trying these tests with money and health companies first[1].\n\nIn the United States, **Texas** made a new law about AI[5]. It says companies must show how their AI works, check for unfairness, and allow safety inspections[5]. This is one of America's strongest state AI laws so far[5]. **Michigan** lawmakers proposed the AI Safety and Security Transparency Act[6]. If passed, it would make big AI companies share safety plans, do yearly outside checks, and protect workers who report problems[6].\n\n**California** is discussing several AI safety bills[6]. One would require people to watch over AI running important things like trains, power systems, and emergency services[6]. Another would protect workers who report risks from powerful AI models[6]. These laws aim to prevent AI accidents in critical areas.\n\nDoctors got important advice about using AI helpers. A study in a medical journal said hospitals must get **clear patient permission** before AI helps with health decisions[5]. The authors warned that vague rules hurt trust between doctors and patients[5]. This matches WDTA's work on health AI safety testing[1].\n\nAt the UN's **AI for Good Summit** in Geneva, experts shared worries about rushing AI into daily life[7]. ITU leader Doreen Bogdan-Martin warned: \"The biggest risk is putting AI everywhere without understanding what it means for people and our planet\"[7]. She asked everyone to help make AI safe and useful for all[7].\n\nCountries in the **BRICS group** (including Brazil, Russia, India, China, and South Africa) asked the United Nations to lead global AI rules[5]. They want fair access to AI technology and ethical oversight[5]. This push for worldwide cooperation comes as AI helpers spread quickly.\n\nBusinesses got reminders about **ethical AI use**. Experts said companies must measure AI trustworthiness and understand where biases come from[2]. They suggested using numbers to track fairness and set clear responsibility lines[2]. Another article advised companies to build privacy protections into AI from the start[4]. This includes using less personal data and strong security locks[4].\n\nAn opinion piece compared using AI in hiring to past tech changes like calculators[8]. The author argued that calling AI assistance \"cheating\" might miss how technology naturally changes work[8]. This debate shows how AI helpers are forcing new conversations about fairness in everyday life."
}
```

WDTA Sets a New Global Benchmark for Single AI Agent ...

What Can Businesses Do About Ethical Dilemmas Posed ...

Responsible AI Implementation: Ethical Considerations for ...

Smart AI and smarter compliance: a blueprint for ...

Latest AI Breakthroughs and News: May, June, July 2025

AI Legislative Update: July 11, 2025

UN summit confronts AI's dawn of wonders and warnings

If AI Can Easily Game Hiring Processes, Maybe It's Time to ...

AI Reporter - July 2025 | Benesch - JDSupra

New safety rules for AI helpers were announced at a United Nations meeting in Geneva. The **World Digital Technology Academy** (WDTA) created special tests to check if single AI agents work safely[1]. These tests are like safety belts for AI helpers in cars, hospitals, factories, and banks[1]. WDTA's leader Yale Li said these rules help "put ethics into every stage of AI's life"[1]. The group is already trying these tests with money and health companies first[1].

In the United States, **Texas** made a new law about AI[5]. It says companies must show how their AI works, check for unfairness, and allow safety inspections[5]. This is one of America's strongest state AI laws so far[5]. **Michigan** lawmakers proposed the AI Safety and Security Transparency Act[6]. If passed, it would make big AI companies share safety plans, do yearly outside checks, and protect workers who report problems[6].

**California** is discussing several AI safety bills[6]. One would require people to watch over AI running important things like trains, power systems, and emergency services[6]. Another would protect workers who report risks from powerful AI models[6]. These laws aim to prevent AI accidents in critical areas.

Doctors got important advice about using AI helpers. A study in a medical journal said hospitals must get **clear patient permission** before AI helps with health decisions[5]. The authors warned that vague rules hurt trust between doctors and patients[5]. This matches WDTA's work on health AI safety testing[1].

At the UN's **AI for Good Summit** in Geneva, experts shared worries about rushing AI into daily life[7]. ITU leader Doreen Bogdan-Martin warned: "The biggest risk is putting AI everywhere without understanding what it means for people and our planet"[7]. She asked everyone to help make AI safe and useful for all[7].

Countries in the **BRICS group** (including Brazil, Russia, India, China, and South Africa) asked the United Nations to lead global AI rules[5]. They want fair access to AI technology and ethical oversight[5]. This push for worldwide cooperation comes as AI helpers spread quickly.

Businesses got reminders about **ethical AI use**. Experts said companies must measure AI trustworthiness and understand where biases come from[2]. They suggested using numbers to track fairness and set clear responsibility lines[2]. Another article advised companies to build privacy protections into AI from the start[4]. This includes using less personal data and strong security locks[4].

An opinion piece compared using AI in hiring to past tech changes like calculators[8]. The author argued that calling AI assistance "cheating" might miss how technology naturally changes work[8]. This debate shows how AI helpers are forcing new conversations about fairness in everyday life.

This week saw major steps in AI safety rules. The **World Digital Technology Academy** launched new safety tests for single AI agents at a UN meeting[1]. These tests act like seatbelts for AI helpers in cars, hospitals, and banks[1]. 

In America, **Texas** passed a big law requiring AI systems to be clear and fair[5]. **Michigan** proposed a new bill making AI companies share safety plans and protect whistleblowers[6]. **California** is debating laws about human control over AI in important systems like power grids[6]. 

Doctors got warned to get **patient permission** before using AI for health decisions[5]. The UN held a special meeting where experts said we must understand AI risks before putting them everywhere[7]. BRICS countries asked the **United Nations** to create fair worldwide AI rules[5].

Ethics & Safety Weekly AI News

Specific Topics