Microsoft has signed a new agreement with the American Federation of Teachers and the United Federation of Teachers to protect student and teacher data. The deal includes ten enforceable rules that stop Microsoft from using school data to train its AI models, with a small exception for safety and security. It also bans AI companions in schools and requires human review for high-risk decisions. Union leader Randi Weingarten called the terms iron-clad because no other group, including the federal government, has stepped up to do this work.
OpenAI CEO Sam Altman is urging the world to slow down the race to develop artificial general intelligence due to serious safety risks. He warned that rushing forward could lead to catastrophic mistakes that we might not be able to fix. Other experts like Andrew Ng agree, stating that scientists are not even sure what they are building. The concern is that top tech companies like Google, Microsoft, and Amazon are racing ahead without enough transparency or accountability.
Samuel Marks, a safety researcher at Anthropic, posted that he believes artificial intelligence could lead to human extinction within the next few years. He explained that developers continue building AI due to commercial incentives and a fear of losing a race to less responsible companies. This warning highlights the ongoing debate about how fast technology should advance and the severe risks involved in creating powerful AI systems.
Meanwhile, Virginia Tech is hosting a New Music + Technology Festival from September 14 to 19 that features animatronics and artificial intelligence. One performance will use interactive robots in a virtual reality experience, while another show will feature an AI named Wormwood that watches and analyzes human dancers in real time. The festival also highlights the Meta settlement, which limits teen social media use and removes push notifications during school hours.
In other developments, the US Treasury Secretary Scott Bessent warned that the United States faces dire consequences if it loses the artificial intelligence race to China. His statement highlights the growing anxiety in Washington about China's rapid technological advances. Additionally, a federal judge blocked Vermont Attorney General Charity Clark from investigating a satirical AI video, ruling it was protected speech under the First Amendment.
Bangladesh is offering free basic AI training for young people in all 64 districts. The two-month program runs from October 1 to November 30, 2026, and costs nothing for participants. Trainees aged 18 to 35 will receive a daily travel allowance of Taka 200 and food during the course, provided they have passed the HSC exam and have basic computer skills.
Key Takeaways
- Microsoft signed an agreement with teachers unions to ban using student data for AI training and prohibit AI companions in schools.
- The deal includes ten enforceable rules requiring human review for high-risk AI decisions in educational settings.
- OpenAI CEO Sam Altman urged researchers to slow down the race for artificial general intelligence due to safety risks.
- Experts warn that top tech companies like Google, Microsoft, and Amazon lack sufficient transparency in their AI development.
- Anthropic researcher Samuel Marks stated AI could lead to human extinction within the next few years.
- Virginia Tech's festival features an AI named Wormwood that analyzes human dancers in real time.
- The Meta settlement limits teen social media use and removes push notifications during school hours.
- US Treasury Secretary Scott Bessent warned losing the AI race to China would have disastrous consequences.
- A federal judge blocked a Vermont probe into a satirical AI video, citing First Amendment protections.
- Bangladesh launched free AI training for youth aged 18 to 35 across all 64 districts starting October 2026.
Microsoft agrees to strict AI privacy rules for schools
Microsoft has signed a new agreement with the American Federation of Teachers and the United Federation of Teachers to protect student and teacher data. The deal includes ten enforceable rules that stop Microsoft from using school data to train its AI models. It also bans AI companions in schools and requires human review for high-risk decisions. Union leader Randi Weingarten called the terms iron-clad because no other group, including the federal government, has stepped up to do this work.
Teachers union and Microsoft reach deal on AI safety
Microsoft and the American Federation of Teachers have reached a deal to ensure safety and privacy when artificial intelligence is used in schools. The agreement contains ten legal protections, including a ban on using student or teacher data to improve AI models, with a small exception for safety and security. Randi Weingarten, president of the union, praised the deal as having real teeth to protect families. She noted that legally enforceable rules are necessary because the federal government has not taken action yet.
Teachers union and Microsoft reach deal on AI safety
Microsoft and the American Federation of Teachers have made a deal regarding safety and privacy when artificial intelligence is used in schools. The agreement has ten legal enforceable protections, one of which stops Microsoft from using teachers or students data to improve AI models. There is an exception for narrow safety and security circumstances. Randi Weingarten praised the agreement as iron-clad and necessary because no one else, including the federal government, has stepped up to do the real work.
Study shows pretraining scores do not guarantee AI quality
A new study finds that language-model checkpoints with the highest pretraining scores are not always the best starting point for future training. Researchers tested a 30 billion parameter mixture-of-experts pipeline and found that better downstream performance requires higher solution density. Solution density means the model retains its performance even when its weights are slightly changed. This discovery challenges the common assumption that the highest-scoring checkpoint will always remain the best choice.
OpenAI urges researchers to slow down the AI race
OpenAI is asking the world to slow down the race to develop artificial general intelligence, or AGI, due to serious safety risks. CEO Sam Altman warned that rushing forward could lead to catastrophic mistakes that we might not be able to fix. Other experts like Andrew Ng agree, stating that scientists are not even sure what they are building or what they are doing. The concern is that top tech companies like Google, Microsoft, and Amazon are racing ahead without enough transparency or accountability.
Virginia Tech festival mixes music, animatronics, and AI
Virginia Tech is hosting a New Music + Technology Festival from September 14 to 19 that features animatronics and artificial intelligence. One performance will use interactive robots in a virtual reality experience called a speculative posthuman cranberry bog. Another show will feature an AI named Wormwood that watches and analyzes human dancers in real time. The festival also highlights the Meta settlement, which limits teen social media use and removes push notifications during school hours.
Bangladesh offers free AI training to youth in all districts
The Department of Youth Development in Bangladesh is organizing free basic AI training for young people in all 64 districts. The two-month program runs from October 1 to November 30, 2026, and costs nothing for participants. Trainees aged 18 to 35 will receive a daily travel allowance of Taka 200 and food during the course. Applicants must have passed the HSC exam and have basic computer skills to join the six-hour daily sessions.
Anthropic researcher says AI could kill all humans soon
Samuel Marks, a safety researcher at Anthropic, posted that he believes artificial intelligence could lead to human extinction within the next few years. He explained that developers continue building AI due to commercial incentives and a fear of losing a race to less responsible companies. Marks argues that the current pace of development ignores the severe risks involved in creating powerful AI systems. This warning highlights the ongoing debate about how fast technology should advance.
Judge blocks Vermont probe of satirical AI video
A federal judge blocked Vermont Attorney General Charity Clark from investigating a satirical AI video called Mark Helps Vermont Take Out The Trash. The judge ruled that the video was ridiculous enough to be protected speech under the First Amendment. This decision stops the enforcement of Vermont Act 75, a law requiring disclaimers on synthetic media used in political ads. The judge found the video qualified for an exemption because it was clearly satire and parody.
US Treasury warns losing AI race to China is disastrous
Treasury Secretary Scott Bessent warned that the United States faces dire consequences if it loses the artificial intelligence race to China. His statement highlights the growing anxiety in Washington about China's rapid technological advances. Bessent emphasized that falling behind in AI could have severe negative impacts on the nation's future. This warning reflects the intense competition between the US and China in the field of artificial intelligence.
Sources
- Microsoft has new AI privacy rules for schools
- AI privacy agreement reached between teachers union and Microsoft
- AI privacy agreement reached between teachers union and Microsoft
- Good Pretraining, Bad SFT: Checkpoint Quality Across the Training Stack
- OpenAI is begging for someone to slow the AI race
- Tech Briefs: Animatronics, AI and more at Virginia Tech’s Music + Technology Festival
- DYD to provide free basic AI training in all 64 districts
- Researcher: People building artificial intelligence believe it could kill all humans
- Federal judge blocks Vermont AG’s probe of ‘Planet Hank’ AI video
- Bessent Warns 'Nothing Would Matter' If China Wins the AI Race
Comments
Please log in to post a comment.