Getting different AIC / BIC values for AR(2) estimation via AutoReg(2) vs ARIMA(2,0,0) through python statsmodels

Question

I am trying to fit an AR(2) model to a data series claims_df['initial claims'] via statsmodels.tsa.ar_model.AutoReg and statsmodels.tsa.arima.model.ARIMA to see if the results are inline. The estimated coefficients are inline but the AIC and BIC values are significantly different for a reason I can't understand. Also, the log-likelihood values are not very different in both models that might explain the difference in AIC / BIC values (sample size is same in both models) considering the formulas for AIC / BIC.

Can someone please explain the difference to me (and which one is correct)?

Below is the code and output snippets for your reference.

File link - https://docs.google.com/spreadsheets/d/1gdV0rBUbGFNHLuAdo88TGx5r5-1s8YSx/edit?usp=sharing&ouid=102682196277760308594&rtpof=true&sd=true

import numpy as np
import pandas as pd
from statsmodels.tsa.arima.model import ARIMA
from statsmodels.tsa.ar_model import AutoReg

claims_df = pd.read_excel('USUNINSCE.xlsx',
                          parse_dates = True,
                          index_col = 0)

# rename columns
claims_df.set_axis(['initial claims', 'dum rec'], axis = 'columns', inplace = True)

# AutoReg
mod = AutoReg(claims_df['initial claims'].values, lags = 2, old_names = False)
res = mod.fit()
print(res.summary())

# ARIMA
arma_mod20 = ARIMA(claims_df['initial claims'].values, order = (2, 0, 0)).fit()
print(arma_mod20.summary())