The New York legislature recently passed the Responsible AI Safety and Education Act (SB6953B) (“RAISE Act”).  The bill awaits signature by New York Governor Kathy Hochul.

Applicability and Relevant Definitions

The RAISE Act applies to “large developers,” which is defined as a person that has trained at least one frontier model and has spent over $100 million in compute costs in aggregate in training frontier models. 

• “Frontier model” means either (1) an artificial intelligence (AI) model trained using greater than 10°26 computational operations (e.g., integer or floating-point operations), the compute cost of which exceeds $100 million; or (2) an AI model produced by applying knowledge distillation to a frontier model, provided that the compute cost for such model produced by applying knowledge distillation exceeds $5 million.
• “Knowledge distillation” is defined as any supervised learning technique that uses a larger AI model or the output of a larger AI model to train a smaller AI model with similar or equivalent capabilities as the larger AI model.

The RAISE Act imposes the following obligations and restrictions on large developers:  

• Prohibition on Frontier Models that Create Unreasonable Risk of Critical Harm: The RAISE Act prohibits large developers from deploying a frontier model if doing so would create an unreasonable risk of “critical harm.”

  • “Critical harm” is defined as the death or serious injury of 100 or more people, or at least $1 billion in damage to rights in money or property, caused or materially enabled by a large developer’s use, storage, or release of a frontier model through (1) the creation or use of a chemical, biological, radiological or nuclear weapon; or (2) an AI model engaging in conduct that (i) acts with no meaningful human intervention and (ii) would, if committed by a human, constitute a crime under the New York Penal Code that requires intent, recklessness, or gross negligence, or the solicitation or aiding and abetting of such a crime.

• Pre-Deployment Documentation and Disclosures: Before deploying a frontier model, large developers must:

  • (1) implement a written safety and security protocol;
  • (2) retain an unredacted copy of the safety and security protocol, including records and dates of any updates or revisions, for as long as the frontier model is deployed plus five years;
  • (3) conspicuously publish a redacted copy of the safety and security protocol and provide a copy of such redacted protocol to the New York Attorney General (“AG”) and the Division of Homeland Security and Emergency Services (“DHS”) (as well as grant the AG access to the unredacted protocol upon request);
  • (4) record and retain for as long as the frontier model is deployed plus five years information on the specific tests and test results used in any assessment of the frontier model that provides sufficient detail for third parties to replicate the testing procedure; and
  • (5) implement appropriate safeguards to prevent unreasonable risk of critical harm posed by the frontier model.

• Safety and Security Protocol Annual Review: A large developer must conduct an annual review of its safety and security protocol to account for any changes to the capabilities of its frontier models and industry best practices and make any necessary modifications to protocol. For material modifications, the large developer must conspicuously publish a copy of such protocol with appropriate redactions (as described above).  
• Reporting Safety Incidents: A large developer must disclose each safety incident affecting a frontier model to the AG and DHS within 72 hours of the large developer learning of the safety incident or facts sufficient to establish a reasonable belief that a safety incident occurred.

  • “Safety incident” is defined as a known incidence of critical harm or one of the following incidents that provides demonstrable evidence of an increased risk of critical harm: (1) a frontier model autonomously engaging in behavior other than at the request of a user; (2) theft, misappropriation, malicious use, inadvertent release, unauthorized access, or escape of the model weights of a frontier model; (3) the critical failure of any technical or administrative controls, including controls limiting the ability to modify a frontier model; or (4) unauthorized use of a frontier model. The disclosure must include (1) the date of the safety incident; (2) the reasons the incident qualifies as a safety incident; and (3) a short and plain statement describing the safety incident.

If enacted, the RAISE Act would take effect 90 days after being signed into law.

The number of school suspensions and exclusions in England has reached its highest level since 2006, Department for Education figures show.

There were 954,952 suspensions in state schools in 2023/24 – a 21% increase on the previous year – while exclusions also rose 16% to 10,885.

While secondary school pupils comprised most suspensions, more than 100,000 were primary age – a number that has grown significantly.

A suspended pupil must stay out of school for a fixed period of up to 45 days per school year, while those excluded are permanently removed. Individual pupils often account for more than one period of suspension.

The government says it is tackling the root causes of poor behaviour and is intensively supporting 500 schools with the worst behaviour.

Persistent disruptive behaviour was the most common reason pupils were sent home, accounting for half of all suspensions and 39% exclusions.

Nearly half of the suspensions were among pupils getting support for special educational needs – who were three times more likely to be suspended than their classmates.

Children on free school meals were also overrepresented, making up a quarter of the school population but 60% of suspensions.

Paul Whiteman, general secretary at school leaders’ union NAHT, said schools alone could not address the causes of poor behaviour.

“Schools have a duty to provide a safe environment for all pupils and only use suspensions and exclusions when other options to ensure this have been exhausted,” he said.

“The reasons for disruptive behaviour often lie beyond the school gates and have their roots in wider challenges, including everything from poverty to access to support with special educational needs and mental ill-health.”

The vast majority of suspensions – nine in 10 – occurred at secondary schools, with Year 9 having the highest rate.

But primary-age suspensions rose too, up 24% on the previous year.

The vast majority (88%) of pupils who were excluded at primary school were getting support for special educational needs, compared with 46% of excluded secondary school pupils.

Research from charity Chance UK, which supports families of excluded children in London, suggests that 90% of children who are excluded at primary school fail to pass GCSE English and maths.

Sophie Schmal, the charity’s director, said Thursday’s figures revealed a “very concerning picture” – particularly the rise in primary school suspensions.

“Early intervention has to mean early. We can’t wait until these children are teenagers to tackle this.”

Sarah – not her real name – is a mum of one in London. Her six-year-old son was suspended several times within his first few weeks at primary school for hitting other pupils and throwing things in class.

She said that even after school staff agreed that her son showed signs of autism, he continued to be sent out of class regularly and suspended, which made him feel “isolated”.

“Since he was three years old, my son has been labelled as the naughty and difficult kid when all he really needed was help,” she said.

“I sought help as soon as I recognised that he needed additional support. But rather than helping me immediately, they waited until it was an emergency.”

Sarah eventually managed to move her son to a different mainstream school where he is getting more support, she said.

Responding to the figures, early education minister Stephen Morgan said the Labour government had “wasted no time in tackling the root causes of poor behaviour”, including offering mental health support in every school and expanding free school meals.

He pointed to its new attendance and behaviour hubs, which will directly support the 500 schools that “need the most help”.

“We’re also continuing to listen to parents as we reform the SEND system, while already putting in place better and earlier support for speech and language needs, ADHD and autism,” Morgan added.

When Lacoochee Elementary School resumes classes in August, principal Latoya Jordan wants teachers to focus more attention on each student’s individual academic needs.

She’s looking at artificial intelligence as a tool they can use to personalize lessons.

“I’m interested to see how it can help,” Jordan said.

Lacoochee is exploring whether to become part of the Pasco County school district’s new AI initiative being offered to 30 campuses in the fall. It’s a test run that two groups — Scholar Education and Khanmigo — have offered the district free of charge to see whether the schools find a longer-term fit for their classes.

Scholar, a state-funded startup that made its debut last year at Pepin Academy and Dayspring Academy, will go into selected elementary schools. Khanmigo, a national model recently highlighted on 60 Minutes, is set for use in some middle and high schools.

“Schools ultimately will decide how they want to use it,” said Monica Ilse, deputy superintendent for academics. “I want to get feedback from teachers and leaders for the future.”

Ilse said she expected the programs might free teachers from some of the more mundane aspects of their jobs, so they can pay closer attention to their students. A recent Gallup poll found teachers who regularly use AI said it saves them about six hours of work weekly, in areas such as writing quizzes and completing paperwork.

Marlee Strawn, cofounder of Scholar Education, introduced her system to the principals of 19 schools during a June 30 video call. The model is tied to Florida’s academic standards, Strawn said, and includes dozens of lessons that teachers can use.

It also allows teachers to craft their own assignments, tapping into the growing body of material being uploaded. The more specific the request, the more fine-tuned the exercises can be. If a student has a strong interest in baseball or ballet, for instance, the AI programming can help develop standards-based tasks on those subjects, she explained.

Perhaps most useful, Strawn told the principals, is the system’s ability to support teachers as they analyze student performance data. It identifies such things as the types of questions students asked and the items they struggled with, and can make suggestions about how to respond.

“The data analytics has been the most helpful for our teachers so far,” she said.

She stressed that Scholar Education protects student data privacy, a common concern among parents and educators, noting the system got a top rating from Common Sense.

School board member Jessica Wright brought up criticisms that AI has proven notoriously error-prone in math.

Strawn said the system has proven helpful when teachers seek to provide real-life examples for math concepts. She did not delve into details about the reliability of AI in calculations and formulas.

Lacoochee principal Jordan wanted to know how well the AI system would interface with other technologies, such as iReady, that schools already use.

“If it works with some of our current systems, that’s an easier way to ease into it, so for teachers it doesn’t become one more thing that you have to do,” Jordan said.

Strawn said the automated bot is a supplement that teachers can integrate with data from other tools to help them identify classroom needs and create the types of differentiated instruction that Jordan and others are looking for.

The middle and high school model, Khanmigo, will focus more on student tutoring, Ilse wrote in an email to principals. It’s designed to “guide students to a deeper understanding of the content and skills mastery,” she explained in the email. As with Scholar, teachers can monitor students’ interactions and step in with one-on-one support as needed, in addition to developing lesson plans and standards-aligned quizzes.

Superintendent John Legg said teachers and schools would not be required to use AI. Legg said he simply wanted to provide options that might help teachers in their jobs. After a year, the district will evaluate whether to continue, most likely with paid services.

While an administrator at Dayspring Academy before his election, Legg wrote a letter of support for Scholar Education’s bid for a $1 million state startup grant, and he also received campaign contributions from some of the group’s leaders. He said he had no personal stake in the organization and was backing a project that might improve education, just as he previously supported Algebra Nation, the University of Florida’s online math tutoring program launched in 2013.