Jindal Steel grew from $750 million to $12 billion

AI agents are prompting businesses to rethink identity and access controls. SailPoint Field CTO Matt Fangman says organizations often discover far more non-human identities than expected: one company found more than 10,000 credentials and AI agents despite technology leaders thinking it had none. Experts recommend assigning each agent separate credentials, limited permissions, an owner and clear records of its activity. Old access rights can expose systems beyond an agent’s intended task; CrowdStrike, microsoft and Palo Alto Networks are among the endpoint security partners that can help block unsafe activity. Dan Moore warns that agents using employees’ regular credentials could cause serious financial or reputational damage.

AI is also moving into consumer services and business operations. Hyundai Home Shopping starts regular broadcasts made entirely by its AI-REAL system this month, with a Jeju Heart Gold kiwi program at 3:20 p.m. on November 16. The system generates both video and voices without a live host or camera filming. In test broadcasts, partners’ sample and video production costs fell by more than 80%, while sales reached 120% of the initial target. Separately, San Francisco voice AI company Bland has released Tofu-1 in a limited enterprise research preview. The model processes incoming and outgoing audio directly during calls, can distinguish speakers and background noise, detect emotional cues, and respond to interruptions. Bland has not disclosed its size, latency or price.

Evidence about AI’s effects on work remains mixed. A Reserve Bank of Australia analysis finds employment among young people in highly AI-exposed fields, including software development and customer service, has declined since late 2022, while employment among older workers in those fields has risen. The author cautions that the analysis measures potential exposure, not actual AI use, and does not establish that AI caused the changes. A separate Business Insider headline says AI could add almost two million jobs net by 2030, but the available article text provides no further details about the estimate or its source. Among surveyed game developers, 40% use generative AI, most often for research, brainstorming, coding or debugging; only 7% use it for player-facing features. Yet 55% say it is harming the industry, with concerns including weakened skills and professional development.

Companies are also weighing AI’s reliability and the skills needed to use it well. A workshop with the Seediq community, organized by the Wikimedia Foundation and Ministry of Education, highlighted how language models can invent information or reproduce inaccurate context. In a demonstration, Grok’s Traditional Chinese-to-Seediq translation mixed dialect features and had grammar and wording problems, showing why subject knowledge matters when checking AI output. Former steel executive Geoff Woods argues that executives should learn to use AI before demanding adoption from employees or linking AI skills to promotions. He says leaders should apply it to strategic thinking, not just routine writing. Woods helped grow Jindal Steel from $750 million to $12 billion and now leads AI Leadership.

AI is also becoming a political issue ahead of the 2028 election. Democratic group Tomorrow Together is preparing to shape how Democrats discuss AI and respond to serious incidents. Some Democrats see public support for guardrails as an opportunity, while others caution that regulation should not undermine the United States’ technology lead over China. Republican Rep. Gabe Evans has likewise backed guardrails while emphasizing the need to protect that edge. In cybersecurity, AI can help teams find hidden software weaknesses, although Scott Miserendino says the backlog of old code to review is finite. Separately, a two-week trial of AI road cameras in Birmingham and Wolverhampton flagged nearly 19,000 possible offences in March, including 17,075 possible seatbelt offences and 1,920 possible mobile-phone offences. No fines or other enforcement were planned during the trial; West Midlands road safety commissioner Mat MacDonald said the aim is to change driver behavior.

Key Takeaways

  • SailPoint Field CTO Matt Fangman says one company uncovered more than 10,000 non-human identities and credentials, despite leaders believing it had none; experts recommend owned, separately credentialed agents with limited permissions and activity records.
  • AI agents using employees’ regular credentials or outdated permissions can reach unintended systems and create financial, reputational and security risks; CrowdStrike, microsoft and Palo Alto Networks can help block unsafe activity.
  • Hyundai Home Shopping begins monthly AI-made broadcasts with a Jeju Heart Gold kiwi program on November 16; test broadcasts cut partners’ sample and video costs by over 80% and reached 120% of initial sales targets.
  • Bland’s Tofu-1 processes call audio directly, supports interruptions and detects speaker and emotional cues; it is in a limited enterprise research preview, with size, latency and price undisclosed.
  • A Reserve Bank of Australia analysis finds young employment has fallen in highly AI-exposed fields since late 2022, but does not establish that AI caused the change. A Business Insider headline projects nearly two million net jobs added by 2030, without further details in the available text.
  • Forty percent of surveyed game developers use generative AI, but 55% say it is harming the industry; only 7% use it for features players see.
  • The Seediq workshop’s Grok demonstration showed translation errors across dialect features, underscoring the need for subject knowledge to assess AI-generated information.
  • Geoff Woods urges executives to learn AI before requiring employee adoption; he helped grow Jindal Steel from $750 million to $12 billion and now leads AI Leadership.
  • Tomorrow Together is preparing to shape Democrats’ AI messaging ahead of 2028, amid debate over guardrails and preserving the U.S. technology lead over China; Republican Rep. Gabe Evans has also called for guardrails.
  • West Midlands AI road cameras flagged nearly 19,000 possible offences in a two-week March trial; no enforcement was planned during the trial, which aimed to encourage safer driving.

AI agents are creating new identity risks for businesses

AI agents can access accounts, buy products, change bookings and use company systems on behalf of people. Experts say businesses need to treat these agents as non-human identities and give them separate credentials, limited permissions and clear records of their actions. Dan Moore warned that an agent using a person’s regular credentials could cause major financial or reputational harm. AI is also helping security teams find hidden software weaknesses, though Scott Miserendino said the backlog of old code to review is finite.

SailPoint warns that unmanaged AI agents create security risks

SailPoint Field CTO Matt Fangman said companies often find far more non-human identities than they expect when they begin adopting AI agents. In one case, a company had more than 10,000 identities, including credentials and AI agents, even though its technology leaders thought it had none. Old credentials and access permissions can let agents reach systems and data beyond their intended tasks. Fangman said businesses should track who owns each agent, what it can access and whether it has approval, while endpoint security partners such as CrowdStrike, Microsoft and Palo Alto Networks can help block unsafe activity.

AI cameras detect nearly 19,000 possible driving offences in two weeks

AI cameras in Birmingham and Wolverhampton detected nearly 19,000 possible offences during a two-week trial in March. The systems recorded 11,045 possible seatbelt offences and 1,163 possible mobile phone offences at Ladywood, plus 6,030 seatbelt offences and 757 mobile phone offences at Fallings Park. West Midlands road safety commissioner Mat MacDonald said images showed dangerous behavior, including a child riding unrestrained on an adult’s lap and drivers using phones. No fines or other enforcement were planned during the trial, which the West Midlands Combined Authority funded with police and fire services. MacDonald said he wants the technology to help change driver behavior, with the eventual goal of no offences being detected.

Many game developers use generative AI, but most see harm to the industry

A report found that 40% of surveyed game developers use generative AI, with older workers more likely to use it than younger ones. The most common uses were research and brainstorming, followed by code generation and debugging. Fewer developers used it for concept art, and only 7% used it for features that players see. Still, 55% said generative AI is having a negative effect on the games industry, while 35% said its impact was mixed and 8% said it was positive. The report also noted concerns that AI could weaken skills and professional development.

Democratic group aims to shape AI debate ahead of 2028

Tomorrow Together, a Democratic political group, is launching an effort to shape how Democrats discuss artificial intelligence with voters. The group supported AI development rules during the midterms and says it is preparing to respond if more serious AI incidents occur. Some Democrats see the issue as a chance to support guardrails, which David Shor said voters from both parties favor. Others warn that rules must not weaken the United States’ technology lead over China. Republican Rep. Gabe Evans has also called for guardrails while stressing that the country must protect its technological edge.

Hyundai Home Shopping begins fully AI-made sales broadcasts

Hyundai Home Shopping will begin regular AI-made sales broadcasts this month on its Hyundai Home Shopping Plus Shop channel. The first program will feature Jeju Heart Gold kiwis at 3:20 p.m. on November 16. Its AI-REAL system creates the full broadcast, including voices and video, without a live host or camera filming. In test broadcasts, partners’ costs for product samples and video production fell by more than 80%, while sales reached 120% of the initial target. The company plans to air at least one AI broadcast a month, starting with agricultural, livestock, and seafood products.

Bland introduces Tofu-1, an AI model that listens and speaks at once

Bland, a San Francisco voice AI company, released Tofu-1 in a research preview for a limited number of enterprises. The model handles incoming and outgoing audio directly during phone calls, without turning speech into text first. It can track speakers, separate voices from background noise, and detect signs of hesitation, frustration, or urgency. Callers can interrupt or correct it while it speaks, and it keeps talking as it works on a request. Bland says the model is aimed at sectors including banking, healthcare, insurance, and government, but has not disclosed its size, latency, or price.

RBA report finds young workers in AI exposed jobs facing declines

An analysis by Mike Major for the Reserve Bank of Australia found that employment among young people in highly AI exposed jobs has fallen since late 2022, while employment among older workers in those fields has risen. Software development and customer service are among the fields highlighted. Major cautioned that the study does not prove AI caused the changes and measures potential exposure rather than companies’ actual use of AI. Changes in the business cycle or remote work after the pandemic may also help explain the figures. The article notes that AI can also support existing work and create new demand.

AI may create nearly two million jobs by 2030, report suggests

A Business Insider headline says AI could add almost two million jobs on net by 2030, even as it may also lead to job losses. The article page identifies Madison Hoff as the reporter covering the labor market and other economic data. The available article text is behind a Business Insider subscriber notice and does not provide further details about the estimate or its source.

Former steel executive urges CEOs to use AI before judging employees

Geoff Woods argues that CEOs and executive teams should learn to use AI themselves before asking employees to adopt it or tying AI skills to raises and promotions. Woods says many companies mandate AI use but do not hold leaders to the same standard, and that executives should use AI for strategic thinking rather than only tasks such as writing emails. He describes starting to explore AI in December 2022 while serving as chief growth officer at Jindal Steel. Woods previously helped grow the steel company from $750 million to $12 billion and now leads AI Leadership, which trains leadership teams to use AI.

Seediq Workshop Highlights Risks of AI in Indigenous Knowledge

The Wikimedia Foundation and Ministry of Education initiated an Indigenous Wikipedia and Artificial Intelligence Workshop with the Seediq community. Participants discussed how large language models can invent false information and build on inaccurate context, making errors harder to spot. In a live demonstration, Grok translated Traditional Chinese into Seediq, but the result mixed features from different dialects. Community members found the translation mostly understandable, with grammar and wording problems. The article says people need strong subject knowledge to recognize when AI-generated information is wrong.

Sources

NOTE:

This news brief was generated using AI technology (including, but not limited to, Google Gemini API, Llama, Grok, and Mistral) from aggregated news articles, with minimal to no human editing/review. It is provided for informational purposes only and may contain inaccuracies or biases. This is not financial, investment, or professional advice. If you have any questions or concerns, please verify all information with the linked original articles in the Sources section below.

AI agents non-human identity management identity and access management cybersecurity least-privilege access AI-generated media AI voice models Bland Tofu-1 AI and employment generative AI in gaming AI reliability AI translation AI leadership AI regulation AI policy AI software vulnerability detection AI road cameras

Comments

Loading...