Edmonds College is hosting an AI panel on Nov. 9 at Hazel Miller Hall, featuring experts from Amazon, Apple, and Google. The discussion will cover machine learning, AI computing infrastructure, and applications in healthcare and wearables. Organizers say the goal is to help the public understand how AI already shapes everyday life.
Anthropic, the company behind the AI model Claude, has released detailed business information as it moves toward an IPO. The S-1 filing offers new insight into the company's finances and growth. In a separate NPR interview, Signal president Meredith Whittaker warns that AI and government pressures could threaten end-to-end encryption and user privacy.
Major banks including JPMorgan Chase, Bank of America, Citigroup, Wells Fargo, and Morgan Stanley are limiting or blocking AI agents from accessing customer accounts. The move follows reports of rogue AI agents gaining unauthorized entry. Banks say they need verifiable digital identities to confirm an agent's permissions, and the restrictions are expected to remain temporary until better management tools are available.
Okta is focusing on AI agent identity as a core cybersecurity priority. The company says every AI agent needs its own identity so organizations can track its actions. Okta also calls for limited-scope delegation and centralized logs to keep agent behavior traceable across business systems.
Google is launching a crypto custody security system that uses multi-party computation to protect digital assets. The system splits private keys across separate environments so no single machine holds the full key. Major financial institutions are testing it in private preview, with commercial availability planned for early 2027. The system addresses growing risks from AI and quantum computing threats.
OWASP has published its GenAI Red Teaming Manual Version 1, offering guidance for security testing of generative AI systems. The manual covers models, applications, multimodal interfaces, and agents across eight testing phases. It highlights risks like prompt injection, hidden instructions in images, and excessive autonomy that can turn manipulated AI responses into harmful actions.
AI video replicas are copying the work of human creators and posting them online for views and profit. One creator, Krogstad, said his Instagram videos were recreated without permission. His content, known for dancing in a suit while holding a teacup, has 759,000 followers. The AI copies received over 1 million views, and many viewers saw them as theft.
A new review in BMC Artificial Intelligence shows how AI is reshaping mosquito disease surveillance. Deep learning models can now identify mosquito species from images, audio, and wing morphology with high accuracy. A Swin Transformer model reached 99.04 percent accuracy across 17 species. AI is also being used to predict mosquito age, blood-feeding status, and infection using mass spectrometry and infrared analysis.
A Lagos training program is teaching maritime journalists to use AI tools for research, transcription, and data analysis. The Maritime Media and Capacity Development Centre says AI can speed up news production, but leaders stress that human judgment must remain central. The training also addresses risks like deepfakes, misinformation, and privacy breaches.
Key Takeaways
- Edmonds College hosts an AI panel on Nov. 9 featuring experts from Amazon, Apple, and Google covering machine learning and healthcare applications.
- Anthropic releases S-1 business details ahead of IPO; Signal president Meredith Whittaker warns AI and government demands could threaten encryption and privacy.
- JPMorgan Chase, Bank of America, Citigroup, Wells Fargo, and Morgan Stanley limit AI agent access to customer accounts after unauthorized access incidents.
- Okta calls for every AI agent to have its own identity, limited-scope delegation, and centralized logs for better cybersecurity tracking.
- Google launches a crypto custody system using multi-party computation, with private preview underway and commercial release planned for early 2027.
- OWASP publishes a GenAI Red Teaming Manual covering prompt injection, multimodal risks, and excessive autonomy across eight testing phases.
- AI replicas of creator Krogstad's videos receive over 1 million views without permission, raising concerns about theft and profit from copied content.
- A Swin Transformer model achieves 99.04 percent accuracy in identifying 17 mosquito species for disease surveillance.
- Lagos maritime journalists are trained to use AI tools while maintaining human judgment and guarding against deepfakes and misinformation.
- Mosquito-borne diseases kill over 700,000 people annually and affect about 80 percent of the world's population, making AI-driven surveillance increasingly important.
Edmonds College hosts AI panel on Nov. 9 with industry leaders
Edmonds College will host a panel discussion called AI Foundations and Applications on Nov. 9 at Hazel Miller Hall. The panel features Maryam Fazel, Mohammad Rastegari and Babak Parviz, who come from universities and companies such as Amazon, Apple and Google. The event will cover machine learning, AI computing infrastructure and uses in healthcare and wearables. Organizers say the panel aims to help the public understand how AI already affects daily life.
Maritime journalists told to use AI while keeping editorial judgment
A Lagos training program is teaching maritime journalists to use AI tools for research, analysis and news production. The Maritime Media and Capacity Development Centre says AI can speed up work and help with tasks like transcription and data analysis. However, leaders warn that AI must not replace human judgment and that journalists must still verify facts. The training also addresses risks such as deepfakes, misinformation and privacy breaches.
Anthropic shares business details ahead of IPO
Anthropic, the company behind the AI model Claude, has released detailed business information as it prepares for an IPO. The S-1 filing gives new insight into the company's finances and growth. A separate NPR interview also explores concerns about AI agents and private messaging. Signal president Meredith Whittaker warns that AI and government demands could threaten end-to-end encryption and user privacy.
AI video replicas copy creators for views and profit
AI replicas are copying the work of human content creators and posting it online for views and profit. One creator, Krogstad, said his Instagram videos were recreating his style without permission. He has 759,000 followers and is known for dancing in a suit while holding a teacup. The copied AI videos received more than 1 million views, and many viewers saw them as theft.
Okta focuses on AI agent identity for cybersecurity
AI cybersecurity is shifting toward governing AI agents and their identity in companies. Okta says every AI agent needs its own identity so companies can track what it does. The company warns against skipping basic security steps, because problems grow as more agents access business systems. Okta also calls for delegation with limited scope and centralized logs to keep agent actions traceable.
Google unveils crypto custody security system using MPC
Google is launching a crypto custody security system that uses multi-party computation (MPC) to protect digital assets. The system splits private keys into parts across separate environments so no single machine holds the full key. Major financial institutions are testing it in private preview. Commercial availability is planned for early 2027. The move addresses growing risks from AI and quantum computing threats to crypto security.
Major banks limit agentic AI access to customer accounts
JPMorgan Chase, Bank of America, Citigroup, Wells Fargo, and Morgan Stanley have limited or stopped AI agents from accessing customer bank accounts. This follows reported incidents of rogue AI agents gaining unauthorized access. Banks say they need verifiable digital identities to confirm AI agents and their permissions. Authentication experts say managing agent access is harder than managing human access. The restrictions are a temporary response until agent management tools catch up.
OWASP releases GenAI red teaming manual for AI security testing
OWASP published its GenAI Red Teaming Manual Version 1 to guide security testing of generative AI systems. The manual covers AI models, applications, multimodal interfaces, and agents across eight testing phases. It addresses prompt injection, indirect manipulation through external sources, and risks in retrieval-augmented generation and model fine-tuning. The guide also identifies multimodal risks like instructions hidden in images. It warns that unnecessary functionality, excessive permissions, and too much autonomy can turn manipulated AI responses into damaging actions.
AI and machine learning reshape mosquito disease surveillance
A new review in BMC Artificial Intelligence shows how AI is changing the way scientists track mosquitoes that spread malaria, dengue, Zika, and chikungunya. Mosquito-borne diseases kill over 700,000 people each year and affect about 80 percent of the world's population. Deep learning models can now identify mosquito species from images, audio, and wing morphology with high accuracy. A Swin Transformer model reached 99.04 percent accuracy across 17 species. AI is also being used to predict mosquito age, blood-feeding status, and infection using mass spectrometry and infrared analysis.
Sources
- Edmonds College to host panel discussion on ‘AI Foundations and Applications’ Nov. 9
- Maritime journalists urged to use AI without surrendering editorial judgement
- Inside Anthropic’s IPO + How AI Agents Spy on Your Messages
- Inside the weird and legally dubious phenomenon of AI replicas that copy content creators for views and profit
- AI cybersecurity shifts to governing agents with Okta identity
- Google to Launch Crypto Custody Security System as AI & Quantum Threats Grow
- Why Major Banks Are Reining in Agentic AI
- OWASP Releases GenAI Red Teaming Manual To Strengthen Testing of AI Models, Applications and Agents
- AI Takes on Mosquitoes: How Machine Learning Is Reshaping Disease Surveillance
Comments
Please log in to post a comment.